Conceptual design of the SPL II

Transcription

1 CERN July 2006 ORGANISATION EUROPÉENNE POUR LA RECHERCHE NUCLÉAIRE CERN EUROPEAN ORGANIZATION FOR NUCLEAR RESEARCH Conceptual design of the SPL II A high-power superconducting H linac at CERN F. Gerigk (Editor), M. Baylac 1, E. Benedico Mora, F. Caspers, S. Chel 2, J.M. Deconto 1, R. Duperrier 2, E. Froidefond 1, R. Garoby, K. Hanke, C. Hill, M. Hori 3, J. Inigo-Golfin, K. Kahle, T. Kroyer, D. Kuechler, J.-B. Lallement, M. Lindroos, A.M. Lombardi, A. López Hernández, M. Magistris, T.K. Meinschad, A. Millich, E. Noah Messomo, C. Pagani 4, V. Palladino 5, M. Paoluzzi, M. Pasini, P. Pierini 4, C. Rossi, J.P. Royer, M. Sanmarti, E. Sargsyan, R. Scrivens, M. Silari, T. Steiner, J. Tückmantel, D. Uriot 2, M. Vretenar GENEVA LPSC Grenoble. 2 CEA Saclay. 3 CERN/Tokyo University. 4 INFN Milan. 5 INFN Naples.

2 CERN 800 copies printed July 2006

3 Abstract An analysis of the revised physics needs and recent progress in the technology of superconducting RF cavities have led to major changes in the specification and in the design for a Superconducting Proton Linac (SPL) at CERN. Compared with the first conceptual design report (CERN ) the beam energy is almost doubled (3.5 GeV instead of 2.2 GeV), while the length of the linac is reduced by 40% and the repetition rate is reduced to 50 Hz. The basic beam power is at a level of 4 5 MW and the approach chosen offers enough margins for upgrades. With this high beam power, the SPL can be the proton driver for an ISOL-type radioactive ion beam facility of the next generation ( EURISOL ), and for a neutrino facility based on superbeam C beta-beam or on muon decay in a storage ring ( neutrino factory ). The SPL can also replace the Linac2 and PS Booster in the low-energy part of the CERN proton accelerator complex, improving significantly the beam performance in terms of brightness and intensity for the benefit of all users including the LHC and its luminosity upgrade. Decommissioned LEP klystrons and RF equipment are used to provide RF power at a frequency of MHz in the lowenergy part of the accelerator. Beyond 90 MeV, the RF frequency is doubled to take advantage of more compact normal-conducting accelerating structures up to an energy of 180 MeV. From there, state-ofthe-art, high-gradient, bulk-niobium superconducting cavities accelerate the beam up to its final energy of 3.5 GeV. The overall design approach is presented, together with the progress that has been achieved since the publication of the first conceptual design report. iii

4

5 Contents List of tables Acronyms and abbreviations vii ix 1 Introduction and executive summary 1 2 Choice of parameters and overall design Neutrino production EURISOL operation Progress in bulk-niobium superconducting cavities since CDR Parameter choices and differences between CDR2 and CDR Applications A neutrino facility at CERN Neutrino physics objectives and facility considerations Radioactive nuclear beam facility EURISOL physics objectives The EURISOL facility The EURISOL design study The SPL as EURISOL driver accelerator LHC upgrade and other CERN applications related to PS injection PS injection and beam dynamics Performance improvement for the LHC Performance improvement for SPS fixed-target physics Performance improvement for the high-intensity beams used by ntof and AD Other advantages and potential improvements Design Main design choices Front end H source and low-energy beam transport (LEBT) Radio frequency quadrupole (RFQ) Low-energy beam chopping Accelerating structures Drift tube linac (DTL), 3 40 MeV Cell-coupled drift tube linac (CCDTL), MeV Side-coupled linac (SCL), MeV Superconducting linac Beam dynamics General approach LEBT (95 kev) v

6 4.4.3 RFQ (95 kev to 3 MeV) Chopper line and matching to the DTL (3 MeV) DTL, CCDTL and SCL (3 MeV to 180 MeV) Superconducting section (180 MeV to 3.5 GeV) End-to-end simulations Error studies Auxiliary systems Transfer line Pulsed operation and servo-system stability Diagnostics Radio frequency sources Cryogenic infrastructure General considerations Heat loads Cryogenic system Open questions Vacuum Layout on site Radiation protection and shielding Civil engineering Cooling and ventilation Primary cooling plant Secondary cooling plant Chilled-water plant Air-conditioning (HVAC) plants Power requirements Electrical infrastructure Access control, safety, interlocks Design options Upgrade options and their limitations Longer beam pulses Higher energy Higher current Failure scenarios Machine and personnel protection Machine operation with loss of non-critical systems References 82 Appendix: SPL parameter list 89 vi

7 Tables 1.1 SPL main characteristics Main layout parameters for all linac sections Main linac parameters and changes from CDR1 to the revised design Tentative parameter list of LHC ultimate beam at low energy in the PS machine Tentative parameter list of a high-intensity SPL beam at low energy in the PS machine Main RFQ characteristics Elements in the chopper line Beam chopper main parameters DTL main parameters CCDTL layout CCDTL main parameters SCL layout SCL main parameters Medium-ˇ multi-cell cavity parameters (bulk-niobium structures only) SNS cavity performances SPL superconducting linac design parameters First estimations on heat losses in the SPL modules assuming a 6% duty cycle Layout of the superconducting sections Lengths in the superconducting linac H single gas stripping cross-section (in cm 2 ) for selected gas types, for ion energies of 100 kev r.m.s. emittance increase between 3 MeV and 180 MeV for initially uniform and Gaussian beams r.m.s. emittance growth and losses for an initially uniform distribution (end-to-end simulation) r.m.s. emittance growth and losses for an initially Gaussian distribution (end-to-end simulation) Total error amplitudes used in the normal-conducting section Sensitivity to errors in the normal-conducting section Total error amplitudes used in the superconducting section r.m.s. emittance growth and losses in the superconducting section General layout of the transfer line RF sources for the SPL Nominal and ultimate SC cavity parameters for the dimensioning of the cryogenic system Estimated heat loads in watts per module assuming a 6% duty cycle Total heat load and installed capacities for a 6% cryo duty cycle Vacuum requirements for the different linac sections Minimum shielding thickness required to reduce the dose rate Maximum ambient dose equivalent rate past the shield vs energy Attenuation properties of two-leg ducts connecting the SPL tunnel to the klystron gallery 65 vii

8 TABLES 5.4 Surface buildings Primary cooling needs and cooling-circuit parameters Secondary cooling circuits Air-conditioning/chilled-water requirements Power requirements for cooling and ventilation Electrical power requirements viii

9 ACRONYMS AND ABBREVIATIONS Acronyms and abbreviations AD Antiproton Decelerator ADS Accelerator Driven Systems APT Accelerator Production of Tritium BARC Bhabha Atomic Research Centre, India BCP Buffered Chemical Polishing BENE Beams for European Neutrino Experiments BINP Budker Institute of Nuclear Physics, Russia BPM Beam Position Monitor CAT Centre for Advanced Technology, India CCC CERN Control Centre CCDTL Cell-Coupled Drift Tube Linac CCL Coupled-Cavity Linac CDR1 Conceptual Design of the SPL, 2000 CDR2 Conceptual Design of the SPL II, 2006 CEA Commissariat pour l Energie Atomique, France CEBAF Continuous Electron Beam Accelerator Facility, USA CNRS Centre National de la Recherche Scientifique CP Charge-conjugation Parity CW Continuous Wave (operation) DC Direct Current DESY Deutsches Elektronen SYnchrotron, Germany DTL Drift Tube Linac EP Electrolytical Polishing ESS European Spallation Source EURISOL EURopean Isotope Separation On-Line HARP Hadron production experiment at CERN HIPPI High Intensity Pulsed Proton Injectors, Europe HOM Higher Order Mode HPR High Pressure Rinsing HVAC Heating Ventilation and Air Conditioning ILC International Linear Collider IN2P3 Institut Nationale de Physique Nucléaire et de Physique des Particules, France INFN Istituto Nazionale di Fisica Nucleare, Italy IPHI Injecteur de Protons de Haute Intensité, France ISOL Isotope Separator On-Line ISOLDE Isotope Separator On Line Experiment at CERN ISR Intersecting Storage Rings ISS International Scoping Study ISTC International Science and Technology Center, Russia ITEP Institute for Theoretical and Experimental Physics, Russia JPARC Japan Proton Accelerator Research Complex, Japan LANL Los Alamos National Laboratory, USA LEBT Low Energy Beam Transport LEP Large Electron Positron collider at CERN LHC Large Hadron Collider at CERN ix

10 ACRONYMS AND ABBREVIATIONS Linac LPSC MEBT MLI MOSFET MPS Nb/Cu NC ntof PLC ppp PPS pps PS PSB RF RFQ RIA RIB r.m.s. RRR SC SCL SEM SNS SPL SPS TERA TESLA TRASCO TTF TTL VNIIEF VNIITF XFEL Linear accelerator Laboratoire de Physique Subatomique et de Cosmologie, France Medium Energy Beam Transport Multi-Layer Insulation Metal Oxide Semiconductor Field Effect Transistor Machine Protection System Niobium on copper Normal-conducting neutron Time of Flight Programmable Logic Controller protons per pulse Personnel Protection System protons per second Proton Synchrotron at CERN PS Booster Radio Frequency Radio Frequency Quadrupole Rare Isotope Accelerator Radioactive Ion Beam root mean square Residual Resistance Ratio (resistivity at 300 K/residual resistivity at operating temperature) SuperConducting Side-Coupled Linac Secondary Electron emission Monitor Spallation Neutron Source, USA Superconducting Proton Linac at CERN Super Proton Synchrotron at CERN TErapia con Radiazioni Adroniche, Italy TeV Energy Superconducting Linear Accelerator project. Lead laboratory: DESY, Germany TRAsmutazione SCOrie, Italy TESLA Test Facility, Germany Transistor Transistor Logic All-Russian Scientific Research Institute of Experimental Physics, Russia All-Russian Scientific Research Institute of Technical Physics, Russia X-ray Free Electron Laser, Europe x

11 Chapter 1 Introduction and executive summary The construction of a Superconducting Proton Linac (SPL) at CERN was initially proposed as an economical means to replace the Proton Synchrotron (PS) injectors (Linac2 and PSB), re-using the valuable asset of RF equipment resulting from the decommissioning of the Large Electron Positron (LEP) collider. When the idea of a neutrino factory based on muon decay was suggested, the SPL, because of its potential for high beam power, became a very attractive solution for the proton driver. The design of the accelerator evolved to match the needs of such a facility, while respecting the constraints of the existing RF equipment. Thereafter a basic scheme for a neutrino factory at CERN was worked upon, until efforts were drastically reduced and most resources were re-oriented towards the Large Hadron Collider (LHC). Only the developments concerning the SPL, in particular its low-energy part, continued to be supported because of their potential application in Linac4, the successor to Linac2. In 2002, an alternative concept for a neutrino facility based on beta-decaying radioactive ions was proposed by P. Zucchelli [1]. The association of such a source of e or anti- e with a neutrino superbeam generating a high flux of or anti- in the same energy range is especially powerful. Assuming that the existing CERN accelerators would be used to accelerate these ions, the optimum energy for the proton driver generating the neutrino superbeam was then determined to be 3:5 GeV [2]. As this energy is also better for the other applications like neutrino factory and PS injector, it has been selected for the updated design of the SPL (see Chapter 2). The basic characteristics for this new design are summarized in Table 1.1. Design details are presented in Chapter 4. Table 1.1: SPL main characteristics Mean beam power 5 MW Beam kinetic energy 3.5 GeV Repetition rate 50 Hz Number of protons per pulse/per second 0:178=8: Pulse length 0.72 ms Average pulse current 40 ma Overall length } 430 m Bunch frequency MHz After chopping. } Pure linac length without debuncher section (see Section 4.5.1). To operate at these increased energy and current levels, more instantaneous RF power is needed than available with the existing LEP klystrons. Combined with the remarkable progress in superconducting RF technology, this led to the decision to use pulsed klystrons and a harmonic of MHz (704.4 MHz) in the superconducting part of the accelerator. As a direct result of this optimization, the second version of the SPL is 340 m shorter than the initial one (430 m instead of 770 m length without debuncher section at the linac end), even though it accelerates up to 3.5 GeV instead of 2.2 GeV. The RF frequency of MHz remains very convenient at low energy, leading to acceptable dimensions, tolerances and gradients in the normal-conducting, low-ˇ accelerating structures like Radio Frequency Quadrupoles (RFQ), Drift Tube Linac (DTL) and Cell-Coupled Drift Tube Linac (CCDTL) up to the energy of 90 MeV. Fourteen LEP klystrons and circulators will be re-used in this part of the linac. In the rest of the SPL, the frequency is doubled to MHz. Normal-conducting SCL structures are employed up to 180 MeV, and multi-cell elliptical superconducting cavities afterwards. It is proposed 1

12 1 INTRODUCTION AND EXECUTIVE SUMMARY to build the first part of the linac, up to 160 MeV (Linac4), in an existing hall close to the PS and to use it as an improved injector for the PSB. The layout of the SPL on the CERN site is similar to that of the previous proposal, except for its reduced length (see Fig. 1.1). The existing transfer tunnels allow for an easy connection to the PS and the ISOLDE/EURISOL experimental area. The accumulator and compressor rings needed for a neutrino factory could be located inside the Intersecting Storage Rings (ISR) tunnel or in its vicinity. 1 40m New fences Access installation 2 Section TYPE 1 95 m SF SU SH SDH Section TYPE 2 Tanks SCX SR 18 kv R=100 m Platform 375 m 150 m m Fig. 1.1: SPL layout on the CERN site The new design and layout of the SPL leaves room for evolution in terms of beam energy and beam power. The space available at the foreseen location allows for a longer accelerator length, and hence a higher beam energy. The infrastructure is already dimensioned for the beam power of 5 MW that is required for an ISOL-type second-generation facility like EURISOL. 2

13 Chapter 2 Choice of parameters and overall design The SPL design in the first conceptual design report (CDR1) [3] was mainly driven by the high-power needs of a CERN-based neutrino factory [4] and the idea to re-use RF equipment [klystrons, waveguides, and superconducting (SC) cavities] from the decommissioned LEP collider at CERN. In the mean time, the needs of the neutrino production schemes have changed and EURISOL became an additional potential high-power user of the SPL. Moreover, the requirements of the CERN proton chain [5, 6] in view of possible LHC luminosity upgrades are more clear [7, 8] and there has been considerable progress in the technology of bulk-niobium SC cavities. The high-power needs of EURISOL and possible neutrino facilities, together with the requirements of LHC upgrade scenarios, led to a revision of the SPL parameters and to a re-design of the linac itself taking into account the latest technical developments in the field of high-intensity proton linacs [9]. 2.1 Neutrino production Three different neutrino production methods are considered to be covered by the SPL (see Section 3.1): a) Superbeam: 4 MW proton beam C accumulator ring C pion production target. b) Beta-beam: 200 kw of the 4 MW proton beam C ISOL-type target C PS, SPS C decay ring. c) Neutrino factory: 4 MW proton beam C accumulator and compressor ring C pion production target C muon cooling and capture channel C muon acceleration C muon decay ring. While the neutrino factory scenario offers the ultimate potential for neutrino physics, a combination of the superbeam and beta-beam scenarios in the same energy range appears as a possible viable alternative. Making use of the CERN infrastructure for the beta-beam scenario, the optimum energy for a CERN-based superbeam facility was determined to be 3.5 GeV [2]. Recent simulation work for options (a) and (c) seems to indicate that the optimum energy for pion production targets is somewhere between 5 GeV and 10 GeV [10, 11]. At these energies the simulations show up to 50% higher transmission through an example muon front-end than for energies around 1 2 GeV. However, for lack of experimental data the computed results in the region of 3 5 GeV are still uncertain and will have to be confirmed by experiments. At present 3.5 GeV seems a reasonable compromise between the size of the linac and the pion production rate, even if it is likely that the optimum energy will be slightly higher. Considering the much more compact design of the revised SPL design, it is possible to increase the energy of the SPL by several GeV if needed (see also Section 6.1). Another commonality of the superbeam and neutrino factory options is the need for an accumulator ring (plus compressor ring in the case of a neutrino factory) to achieve the required time structure. The base line scenario for CDR1 and this second design (CDR2) is the assumption to use the ISR tunnel for this purpose. In CDR1 the large circumference ( 1 km) of the ISR tunnel was needed to limit the number of injection turns. Owing to the long beam pulse length (2.2 ms), however, 660 injection turns were still needed to fill the accumulator/compressor ring at a repetition frequency of 75 Hz. At this high number of turns the heating of the H injection foil and possible beam instabilities during the injection period may develop into intensity limitations for the whole proton driver complex. One goal in the redesign of the SPL was to remove this potential bottleneck and at the same time to reduce the size of the accumulator/compressor ring. In the new scheme for neutrino operation of the SPL, the beam pulse length is reduced to 0.57 ms at 3.5 GeV and would require only 176 injection turns into an ISR-sized ring instead of 660 as assumed in the previous design (see Fig. 2.1). The shorter pulse length opens the possibility to reduce the size of the accumulator/compressor ring (and by doing so to reduce the cost of lattice) while still keeping a reduced number of injection turns. A reduction in size of 50% would 3

14 2 CHOICE OF PARAMETERS AND OVERALL DESIGN 6 #?> > 6 A@B 0(' 23 $! 0('( :) )<; + (= C D E$ F0$G H 0 8(I :) ( 2 C/ 76 # 8 ('29! "$# % &'() * "$+ %-,. /102' 34 $5! Fig. 2.1: Possible timing scheme for neutrino operation assuming an accumulator and compressor ring with a 44 MHz RF system in the ISR tunnel require 360 injection turns and would result in a ring that could be built either in a new tunnel inside the ISR or at a different, more convenient location on the CERN site. 2.2 EURISOL operation Two types of target are foreseen in the reference scenario of the EURISOL design study [12] (see Section 3.2). For the time being, both the high-power (5 MW) and the low-power (100 kw) targets assume a continuous-wave (CW) proton beam at 1 GeV. The transition to pulsed beams will in both cases reduce the lifetime of the targets because of the increase of thermal stresses. In the case of the low-power targets, first studies [13] indicate that for a pulsed machine the pulse lengths should be of the order of 1 ms and the repetition rate should be 50 Hz (or higher) to minimize the instantaneous thermal load in the target. So far the various cross-sections for fragmentation and/or spallation in the low-power target do not indicate a clear advantage or disadvantage of using a 3.5 GeV beam rather than a 1 GeV beam. High-power target studies [14] for CW beams at different energies show that the spallation efficiency reaches its maximum between 1 GeV and 2 GeV, though there is little difference in neutron flux for a 1 GeV or 3.5 GeV beam. At higher energies the heat deposition in the target will be more evenly distributed and there is a significant reduction in the power deposited in the window. On the other hand, more primary protons will escape at higher energies if one assumes an equal target length for 1 GeV and 3.5 GeV beams. The impact of pulsed operation on EURISOL high-power targets has not yet been studied in detail. However, similar challenges are faced by the Spallation Neutron Source (SNS) [15] or in the studies for the European Spallation Source (ESS) [16]. A possible timing scheme to produce a 5 MW beam for EURISOL is depicted in Fig In this scenario, the low-energy beam chopper creates a 0:1 ms gap in the bunch train to switch the beam from the high-power to a low-power target. A second possible way to share the beam within one pulse is to use partial resonant laser stripping to shave off a fraction of the beam over the whole pulse length, which can then be diverted to one (or more) different users. Given that the R&D on laser stripping is showing success, this method would relax the instantaneous thermal shocks on the low-power targets and avoid wasting RF power during the 0.1 ms gap in the bunch train. 4

15 K 2.3 PROGRESS IN BULK-NIOBIUM SUPERCONDUCTING CAVITIES SINCE CDR1 m T n m P$VopR$Sq`rSkn$R` gk$kzstu`vr4skn$r4` X KZYM[ \ T ]5^UR_a`5T V ]<_b_c R$d J$KMLONQP$RSUT V W K2e f gbk2e h g K e h X `jikym[l SkRYjV$w4RW=x4by_ m V PP$RSq`?VQ[oZT `v_ m ` m RQx$Rr)Yzx$R4ÙopR4R]{`rSknR4`?[ Fig. 2.2: Possible timing scheme for EURISOL operation 2.3 Progress in bulk-niobium superconducting cavities since CDR1 In CDR1 [3], the layout of the superconducting section was based on re-using decommissioned ˇ D 1 elliptical cavities which became available at the end of 2000 after the shutdown of LEP. Since the development of the LEP cavities, significant progress has been made in the field of superconducting RF [17]. In particular, starting from the success of the procedures introduced for the mass production of LEP and CEBAF cavities, the TTF/TESLA Collaboration [18, 19] has in the past decade pushed the state-ofthe-art superconducting RF technology of bulk-niobium cavities to new levels in terms of achievable accelerating gradients [20]. Moreover, the bulk-niobium technology has been successfully applied to the design of elliptical cavities for high-current proton accelerators for various applications, ranging from the Accelerator Production of Tritium (APT) at LANL [21] to nuclear waste transmutation in Accelerator Driven Systems (ADS) [22, 23, 24], neutron spallation sources (SNS) [25], and radioactive beam production (RIA) [26]. To cover the energy range from 120 MeV to 1080 MeV it was foreseen in CDR1 to develop three new families of medium-ˇ elliptical multi-cell superconducting cavities using the same technology (niobium sputtered on copper, Nb/Cu) that was used for the LEP cavities. At the time this technology was preferred to bulk-niobium cavities because it promised higher stiffness for the operation in pulsed mode and because it was a well-established (and tested) technology at CERN. For the revised design, presented in this report, two families of MHz bulk-niobium elliptical multi-cell cavities are foreseen for the superconducting portion of the linac. The technology of niobium sputtering on copper substrates which has also evolved considerably since its use in LEP [27] in principle has the perspective of achieving similar RF performances and cannot be excluded in terms of potential RF performances. However, the Nb/Cu technology requires more R&D activities focused on its adaptation and optimization to the reduced-ˇ elliptical resonators, as clearly indicated by the extensive experience gained at CERN on several prototypes [28]. Moreover, for the medium-ˇ structures the low-pressure operation at 2 K is likely to be needed both to reach low values for the surface resistance and to ensure a stable environment for operation. This need softens the perspectives of cost reduction of the Nb/Cu technology, usually dominated by the simplicity of operation at 4.2 K. In addition to achieving high gradients it was also demonstrated that cavity vibrations and Lorentzforce detuning during pulsed operation can be reduced via stiffening and controlled with fast piezo tuners (e.g. [15, 29, 30]). As discussed in Section , the SPL superconducting linac design is based on achieving E peak D 50 MV/m and B peak D 100 mt on the cavity surface, values which are consistent with the performances of bulk-niobium structures reported by several laboratories [20, 31]. In the case of the SPL, two families (ˇ D 0:65 and ˇ D 1:0) of five-cell bulk-niobium elliptical cavities have been chosen to cover the energy range from 180 MeV to 3.5 GeV. Using the above-mentioned peak surface fields and the ratios of peak field over accelerating field, we assume accelerating gradients of 19 MV/m and 25 MV/m for ˇ D 0:65 and ˇ D 1 cavities. 5

16 2 CHOICE OF PARAMETERS AND OVERALL DESIGN 2.4 Parameter choices and differences between CDR2 and CDR1 The current planning for the SPL foresees as a first step the construction of Linac4 [32, 33, 34, 35], which will replace the present Linac2 at CERN and which is designed to operate at the SPL duty cycle. For the construction of the SPL, Linac4 will be relocated in the SPL tunnel to serve as normal-conducting front end. The output beam energy of 160 MeV for Linac4 is chosen to reduce the space-charge forces at injection into the PSB by 50% compared with Linac2, and should thus allow the beam brightness out of the PSB to increase by a factor of 2 [36]. Since the foreseen ˇ D 0:65 cavities have only poor acceleration efficiency at 160 MeV, it was decided to increase the normal-conducting section up to 180 MeV before starting with the superconducting part of the linac. The two families of SC cavities then accelerate the beam up to its final energy of 3.5 GeV. Figure 2.3 shows a block diagram of the linac and Table 2.1 lists the main layout parameters. ŠŸž. (š. (š. (š Ö± ² (š p Š Dš p j q 5 <} ~ pƒo j ˆ p ŠŒ - 1Ž ) < - 1Ž y Ž Šqœ Ÿžk j C œ C µˆ p ž 1 ªj«U ˆ ŠMž³œ (š Fig. 2.3: Schematic layout of the linac Table 2.1: Main layout parameters for all linac sections Section Energy No. of No. of Peak RF No. of No. of Length range cavities accel. energy LEP new [MeV] gaps [MW] klystrons klystrons } [m] Source, LEBT RFQ :0 1 6 Chopper line :1 3:7 DTL :8 5 13:6 CCDTL :4 8 25:5 SCL :1 5 34:9 ˇ D 0: : ˇ D 1: : Total 161: LEP klystrons: MHz, 1 MW. } New klystrons: MHz, 4 5 MW. The energy of 3.5 GeV is chosen as a compromise between the needs of a neutrino production target and the needs of a CERN-based EURISOL proton driver. Because of the higher beam energy and the higher mean current during the pulse, the pulse length could be reduced from 2.2 ms in CDR1 to 0.57 ms in CDR2 (comparing the neutrino operation mode). The shorter pulse length eases the injection into subsequent rings and also relaxes the requirements on the H source. The average pulse current of 40 ma in the neutrino operation corresponds to a peak bunch current of 64 ma after the chopper line and to 80 ma from the source. Beam dynamics studies showed that the space-charge 6

17 2.4 PARAMETER CHOICES AND DIFFERENCES BETWEEN CDR2 AND CDR1 levels resulting from these currents are still acceptable for the low-energy part of the linac, yielding only moderate emittance growth in the area of the chopper line and the subsequent DTL. Since the klystrons are designed for operation at a 10% duty cycle, a wide range of machine modifications becomes possible without fundamentally changing the linac design: lower peak currents with increased pulse length (in case the H source is not able to deliver the required peak design currents), upgrades in beam power by lengthening the beam pulse, delivering two (or more) beam pulses within the same linac pulse (supply of different users at 50 Hz repetition rate), etc. It is also conceivable to add a second low-energy section (source, RFQ, chopper) with a proton source to add more flexibility (in terms of pulse length and peak current) for users which do not need H injection. With the compact layout of the SC section, energy upgrades become feasible at moderate cost (see Section 6.1). Before every beam pulse, the SC cavities have to be filled with energy in order to build up the required accelerating voltage. After the beam pulse when the klystrons are switched off, the stored energy decays and is reflected into the RF loads. During the filling and decaying process a large amount of RF power is reflected into the RF loads without being used for acceleration. The smaller SC cavities at MHz at a cryogenic temperature of 2 K have reduced considerably the cavity filling (and decay) time. Using higher currents in the SC cavities increases the cavity beam loading and therefore reduces the loaded Q and thus lowers even further the filling time constant. In the normal-conducting section, higher currents increase the RF efficiency on account of a higher ratio of beam power over dissipated power. Taking all of these effects together, the average power consumption of the RF system for neutrino (4 MW) operation could be reduced from 24 MW to 17 MW. At the same time, thanks to the lower cryogenics duty cycle, the power needed for the cryo-compressors could be reduced by more than 50%. The disadvantage of using higher currents is that higher RF peak power needs to be installed. However, considering that a) thanks to the development of fast phase and amplitude shifters (see Section ), tetrodes are no longer needed for the medium-ˇ SC section, and that b) purpose-built 704 MHz 5 MW pulsed klystrons are foreseen, rather than 352 MHz 1 MW CW klystrons from LEP, the overall number of RF systems could even be reduced. A full list of the general SPL operational parameters is given in Table 2.2 together with the major changes with respect to CDR1. Table 2.2: Main linac parameters and changes from CDR1 to the revised design (CDR2) CDR1 CDR2 (neutrino) CDR2 (ISOL) Energy [GeV] Length [m] Average beam power [MW] Average RF power [MW] Average cryogenics power [MW] Repetition rate [Hz] Beam pulse length [ms] Average pulse current (after chopping) [ma] Peak bunch current (after 3 MeV) [ma] Beam duty cycle (after chopping) [%] Injection turns (into ISR) Peak RF power [MW] No. of MHz tetrodes (0.1 MW) No. of MHz klystrons (1 MW) No. of MHz klystrons (5 MW) Cryo-temperature [K] Without 30% margin for Lorentz detuning. 7

18 2 CHOICE OF PARAMETERS AND OVERALL DESIGN The main differences with respect to CDR1 can be summarized as follows: energy increase by 60% and reduction in linac length by 45%; reduction in pulse length by almost a factor of 4; 30% lower RF power consumption, > 50% lower cryogenics power consumption; higher RF peak power, from 32 MW to 162 MW; more flexible design in terms of pulse length, beam energy, pulse current; more design flexibility for accumulator/compressor rings; only two families of SC cavities instead of four; bulk-niobium SC cavities instead of Nb/Cu; higher RF frequency (704.4 MHz) from 90 MeV onwards; only one RFQ (95 kv 3 MeV, before the chopper line) instead of two RFQs (45 kv 3 MeV and 3 MeV 7 MeV, before and after the chopper line); higher transition energy between the normal and superconducting part of the linac, 180 MeV instead of 120 MeV. 8

19 Chapter 3 Applications 3.1 A neutrino facility at CERN Recent ideas and developments in the USA [37], in the UK and in Japan [38] have made obsolete the initial CERN scenario. All schemes are still based on the use of muons produced by a high-power proton beam hitting a target, but promising new techniques are nowadays considered for capture and bunch rotation of the muon beam, as well as for muon acceleration. The analysis of these options is the subject of the International Scoping Study [39], which should end in 2006 with the selection of preferred scheme(s) and their comparative assessments. The main evolution for the proton driver concerns the optimum energy, which is now estimated to be between 5 GeV and 10 GeV. The results of the HARP experiment [40] which are expected during the first half of 2006 will be crucial to refine this estimate. A superconducting proton linac like the SPL remains an excellent solution for efficiently providing a high-power proton beam at such energies. The specific time structure for the beam hitting the target will have to be obtained by an adequate design of the necessary ring(s) accumulating the linac burst and compressing the bunches. The details of the linac pulse sequence will also be tailored to these needs Neutrino physics objectives and facility considerations The International Scoping Study (ISS) [39], currently being carried out by BENE (Beams for European Neutrino Experiments) [41], defines the physics potential of accelerator-based neutrino facilities as follows [41, 39]. Neutrino oscillations are now proven to exist and their mixing angles and mass differences should now be precisely measured. They may also permit the first observation of the Charge-conjugation Parity (CP) symmetry in the leptonic sector and provide support to the concept of a CP-violating mechanism (leptogenesis) by which the antimatter was removed from the early Universe. Massive neutrinos may also play an important role in the formation of large-scale structures in the Universe. Precision measurements of neutrino oscillations are required to develop a complete description of the properties of the neutrino and to elucidate the mechanisms that give rise to neutrino mixing and neutrino mass. Two facilities are under consideration. A neutrino factory, based on a high-brilliance muon beam aiming high-energy electron neutrinos (up to GeV) at magnetic detectors km away. This facility is the preferred tool to perform high-precision unambiguous measurements of oscillation parameters, neutrino mass hierarchy, CP violation and tests of universality in the neutrino sector. A neutrino beta-beam facility [1] in combination with a conventional muon-neutrino beam (superbeam) [2] of the same energy. The beta-decay of specific ions would provide a very clean beam of electron (anti)neutrinos up to 2 GeV. The large detectors that are required may be the same as those needed for proton decay and astrophysical neutrinos. Together with a superbeam, beta-beams could provide interesting sensitivity in the search for leptonic CP and T violation. Both options would extend considerably the knowledge of neutrino masses and mixings and offer appreciable synergies with other fields within and outside particle physics. Precision measurements of neutrino mixing angles and mass hierarchy provide a very strong physics case, independent of the explorations at high-energy colliders. 9

20 3 APPLICATIONS 3.2 Radioactive nuclear beam facility Radioactive Ion Beams (RIBs) can be produced either by the ISOL (Isotope Separator On-Line) method or by the complementary in-flight method. The proposed new infrastructure EURISOL [12], the nextgeneration ISOL RIB facility in Europe, aims at the provision of high-intensity beams of radioactive nuclei with variable energy, from a few kev to greater than 100 MeV per nucleon. These beams of exotic ions will be orders of magnitude more intense than those currently available. EURISOL will provide a facility for research that addresses the major challenge of the fundamental understanding of nuclear structure in terms of the underlying many-body interactions between hadrons. Descriptions of nuclei having more than a few nucleons are semi-phenomenological in origin and cannot be reliably applied to nuclei far from stability. It is therefore crucial to measure the properties of nuclei at the extremes of stability such as the evolution of shell structure, T = 0 and T = 1 pairing, and collective phenomena such as halo effects and pygmy resonances EURISOL physics objectives EURISOL also aims at understanding the Universe through its history of stellar activity and galaxy formation where nuclear reactions play essential roles. In particular, in the violent maelstrom of explosive processes such as novae, X-ray bursters and supernovae, the heavy elements are made in complex networks of reactions (r- and rp-processes) on unstable nuclei and beta decays. To understand these processes quantitatively and identify the astronomical sites where they occur requires a wealth of information on unstable nuclei. EURISOL, with its broad range of beams, will allow us to study many of the key reactions. Further applications of EURISOL to fundamental tests of the Standard Model, to the application of unique probes for surface science and condensed matter studies, and to other fields can be found in the report for EURISOL prepared within the RTD contract in the Fifth Framework programme [42]. EURISOL will have applications that benefit society in many different areas and have strong impact on other fields of science, as illustrated in the NuPECC reports of the Working Groups Impact, Applications and Interactions of Nuclear Science [43]. A capability of EURISOL that has been recently identified is the beta-beam facility, which will address the main challenges of neutrino physics: the measurement of the mixing angle 13 of the Maki Nakagawa Sakata Pontecorvo matrix (a fundamental parameter of the Standard Model of interactions) and the search for CP violation in the lepton sector. These measurements, which require intense neutrino beams of defined flavour accompanied by large detectors located at a distance corresponding to the maximum of the observed oscillations, will provide the key ingredient of leptogenesis scenarios for explaining the matter antimatter asymmetry in the Universe The EURISOL facility Historically, nuclear science has progressed through the application of numerous complementary investigative techniques to a broad range of nuclear species. EURISOL will allow European scientists to apply this heritage to vast hitherto unexplored regions of the nuclear chart made accessible through the development and implementation of challenging novel techniques. The EURISOL facility will provide a uniquely broad range of radioactive ion beams with intensities on average three orders of magnitude larger than current ISOL installations and with unrivalled flexibility in final energy between a few ev and 100 MeV/nucleon. In this energy domain, unmatched beam quality will be offered so that our knowledge of nuclei far out towards the drip-lines can approach the precision of that of the stable nuclei. This advancement of performance compared to present-day ISOL facilities will be accomplished through major technical breakthroughs concerning several key components. Unequalled primary beam power of 5 MW protons of up to a few GeV energy, as well as heavy ions of A/q = 2 up to 500 A MeV and possibly also ions with A/q = 3 or larger up to 100 A MeV. 10

21 3.2 RADIOACTIVE NUCLEAR BEAM FACILITY A choice of newly designed production targets: (a) a direct target accepting up to 100 kw beam intensity; (b) a multimegawatt target station based on Hg liquid proton-to-neutron converter and UCx production target; a high-z solid converter will also be investigated. A CW linear post-accelerator for heavy ions delivering a continuous choice of energies up to 100 MeV/nucleon. This will also permit fragmentation of the RIBs to produce even more exotic, neutron-rich nuclei. Based on the EURISOL RTD report, one may expect final intensities of approximately 10 9 pps of 11 Be or pps of 132 Sn after post-acceleration, for example The EURISOL design study The European Union is supporting a design study for EURISOL within the Sixth Framework programme. The study started on 1 February 2005 and will run until the end of It will result in a technical design report for the facility and, in addition, a conceptual design report will be delivered for the betabeam facility. In total 21 research institutes and universities are participating in the study, which has a total budget of C32 million with C9 million coming from the European Union. In addition, 21 institutes and universities are participating as contributors The SPL as EURISOL driver accelerator For EURISOL it is proposed to operate four target stations in parallel: three low-power (< 200 kw), direct proton-target stations and one high-power (5 MW) two-stage spallation neutron target station. The primary proton beam incident on such a facility will be required to meet constraints imposed by the characterization and operation of these target stations, their lifetime and the sharing of the primary beam between targets. The SPL provides many features that are compatible with the EURISOL primary proton beam. Important considerations when choosing beam parameters are target stability and lifetime. All four targets would benefit from continuous beam operation, with smooth, constant proton-beam currents. In practice, proton beams are pulsed and deposit energy within the target volume in short bursts leading to thermal shocks and shock-wave effects which affect the target lifetime. Targets can be heated to mitigate the damage caused by thermal shocks. A drop in incident beam power can then be compensated for by an increase in target heating, although this can only be done on timescales much larger than the pulse length. It is important to design a system where the deposition of energy in the targets is as constant in time as possible. Experience of operating targets at ISOLDE under pulsed proton-beam conditions can help to place limits on the type of pulse that could be acceptable. The following arguments for a 1 ms pulse width were arrived at independently for the development of EURISOL targets: suited for measurements of release properties of targets, reduced thermal stress on the 5 MW target, equivalent to the amount of energy delivered currently at ISOLDE over 920 ns, suited for operation as a 5 kw average beam if necessary, with one pulse every 1 s. At any given time the high-power target station will be operated continuously in parallel with one or two of the low-power targets. This is the most likely scenario since the low-power targets need to be changed every three weeks, so that every week one of the three low-power target stations will be out of action whilst the used target is removed and a new target is installed. The most convenient way of switching between the high-power target station and one of the low-power target stations involves switching within a given pulse. Assuming an SPL pulse of 0.82 ms, the high-power target would receive a bunch train for 0.71 ms, which would be followed by a gap of 0.1 ms during which the pulse would 11

22 3 APPLICATIONS be switched to a low-power target for the remaining 14 s. This implies a change in the microstructure originally planned for SPL, where a pulse consisted of bunches organized in 2: cycles of eight bunches, with five filled bunches and three empty bunches. For EURISOL, the modified pulse structure would consist of 2: filled bunches sent to the high-power target, followed by 3:510 4 empty bunches during which the beam is switched to the low-power target, and finally filled bunches sent to the low-power target as depicted in Fig The operation procedure for targets will include an initial period during which the target is exposed to short bursts of a fraction of the total proton-beam power for which it is rated (Fig. 3.1). This initial period is required for target characterization where release and yield measurements for various isotopes are undertaken and the burst width and time between consecutive bursts would be determined by the release, effusion, diffusion and half-life of the particular isotope to be analysed. Target characterization can then be followed by target warming-up with a ramping up of the beam power to its final nominal value for operation. Some flexibility of the primary proton-beam power will therefore be required. Under the current scheme, the direct (low-power) target will receive a 14 s pulse with an associated instantaneous power deposition of 143 MW. This could be reduced by introducing partial laser stripping of the long pulse to feed a fraction of the 0.71 s pulse to the direct targets, leading to the same average power but a factor 50 lower peak power deposition. P characterisation warming up operation t Fig. 3.1: Operation of a target at EURISOL Lowering of the beam power can be achieved, for example, by reducing the number of pulses incident on the target, by decreasing the pulse length, by altering the pulse microstructure, by reducing the proton energy, or by reducing the beam current. A reduction in the beam current (by a factor 10) is the most favourable scenario for targets, although this is difficult to achieve in practice. It is more likely that either the length of the pulse will be adapted or the low-energy beam chopper will be used to reduce the average current during the pulse. 3.3 LHC upgrade and other CERN applications related to PS injection PS injection and beam dynamics Modifications limited to the hardware handling the injection process are sufficient for the PS to benefit from the SPL beam. These have been described in a previous report [44] and mostly involve removing the present 1.4 GeV fast injection system and replacing it with a charge-exchange injection set-up in straight section 78. The SPL will supply H to the PS with a repetition rate of 1 Hz, meaning that only one pulse in 50 is sent to the PS for the needs of the CERN complex of high-energy accelerators. The duration of these pulses and their fine time structure are completely defined by the PS needs for each specific cycle in the PS super-cycle. Most SPL beam pulses therefore remain available for high-power users like EURISOL or a neutrino facility. As outlined in Chapter 2, the low-energy beam chopper can be 12

23 3.3 LHC UPGRADE AND OTHER CERN APPLICATIONS RELATED TO PS INJECTION used to modify the time structure of the linac pulses and it thus becomes possible to supply pulses to the PS with up to 0:7 ms length and average pulse currents of up to 40 ma. Furthermore the chopper can be used to cut a gap into the otherwise continuous bunch train long enough to switch the highenergy beam to different users (e.g., the existing ISOLDE facility at CERN, compare also the foreseen EURISOL operation scheme in Fig. 2.2). This means that more than one user can be supplied with a 50 Hz repetition rate and with different beam power levels. The main limitation of the PS machine for high-intensity, high-density beams comes from spacecharge detuning at low energy [45], given by Q x;y / I p " x;yˇ 2 (3.1) where Q x;y is the incoherent (self-field) detuning in the horizontal (x) and vertical (y) planes, I p is the bunch peak current, " x;y are the normalized emittances, the ratio I p=" x;y is the beam brightness, and ˇ and are the usual relativistic factors. As an injector for the PS machine, replacing the present Linac2 and PS Booster, the SPL increases the minimum energy in the PS from 1.4 GeV to 3.5 GeV, which reduces the space-charge detuning Q x;y by a factor of 3.8 (instead of 1.9 for CDR1 [3]). The beam brightness I p =" x;y can then be quadrupled while keeping the same Q x;y. This means that the space-charge detuning at injection is no longer a limitation for the acceleration of high-intensity, highbrightness beams in the PS. H injection allows control (by painting) of the particle density in the transverse phase planes. The fast chopper at the SPL front end gives the possibility to optimize the population of the longitudinal phase plane, and in particular to use the harmonic number and the number of bunches that are best matched for each specific PS beam user Performance improvement for the LHC A higher space-charge tune shift than in the nominal scheme using the PSB [46] can be tolerated because the beam stays a much shorter time at the injection energy (the ultimate intensity is accumulated in 30 turns or 60 ms instead of 1.2 s when using the PSB). Combined with the gain resulting from the higher injection energy (see Section 3.3.1), we extrapolate from a previous study [45] that the ultimate intensity for the LHC can probably be achieved with a transverse emittance smaller than 1 mm mrad (rather than the present 3 mm mrad). The brilliance, at the PS exit, will then be up to four times higher than nominally specified [46], which makes it compatible with all foreseen scenarios of LHC luminosity upgrade, although other limitations in the PS are likely to become dominant (collective effects like instabilities, to be addressed with aggressive damping systems). The main beam parameters at the injection energy in the PS are summarized in Table 3.1. Table 3.1: Tentative parameter list of LHC ultimate beam at low energy in the PS machine at W D 3:5 GeV, V RF D 50 kv, h D 21 (only 18 buckets filled), N t D.h 3/N b D 1: protons per pulse, assuming 8% transmission to the LHC N b [10 11 ] " l [ev s] b [ns] " x [ mm mrad] " y [ mm mrad] Q x Q y The fast PS filling process also has the advantage of reducing the PS cycle length from 3.6 s to 2.4 s, which shortens the injection flat-top in the SPS from 7.2 s to 4.8 s and reduces the LHC filling time accordingly. 13

24 3 APPLICATIONS Performance improvement for SPS fixed-target physics With the same arguments as for the LHC beam (see previous section), we estimate that the transverse emittances of a PS beam of, for example, protons will be 4 mm mrad in the horizontal and 6 mm mrad in the vertical plane (at present 17 mm mrad and 7 mm mrad, respectively) while keeping the tune shifts around 0.1. This will eliminate one of the major limitations for achieving a higher intensity (e.g., for the CNGS beam) in the SPS machine which comes from the limited SPS transverse acceptances at injection energy. The estimated parameters for this beam at injection energy in the PS are given in Table 3.2. Studies aimed at a new type of continuous-transfer extraction to reduce the high losses currently related to this operation have been fully demonstrated and are now giving the possibility to increase beam intensity without irradiating the accelerators equipment more. The complete operational implementation is expected to take place before Table 3.2: Tentative parameter list of a high-intensity SPL beam at low energy in the PS machine at W D 3:5 GeV, V RF D 50 kv, h D 16, N t D hn b D protons per pulse N b [10 12 ] " l [ev s] b [ns] " x [ mm mrad] " y [ mm mrad] Q x Q y Performance improvement for the high-intensity beams used by ntof and AD Keeping acceptable tune shifts (0.24 in both planes), the ntof bunch intensity can be increased to protons per pulse (ppp) (0: at present) by adjusting the horizontal and vertical emittances to 6 m and 8 m, respectively. Note that for this beam the longitudinal emittance should not be made smaller than 2.8 ev s to avoid beam break-up instabilities at transition and not higher than 3 ev s to maintain a short bunch at extraction. The intensity of the antiproton beam could also be increased to ppp (at present: 1: ) Other advantages and potential improvements The high proton flux and the fast cycling rate of the SPL decouples the PS and the ISOLDE operations. The PS does not compete with ISOLDE for particles from the injectors, in contrast to the present situation with the PSB. Using the SPL as the injector, the PS is no longer linked to the present linac and PSB repetition rate (i.e., multiples of 1.2 s), and in principle magnetic cycles of any length can be envisaged. This should bring some gain in the duty factor of the PS by reducing the dead time between cycles. However, a significant improvement can only be achieved if the db=dt in the bending magnets is noticeably increased, which implies major hardware changes. The most interesting solution to get the full benefit of the SPL for high-energy physics is to equip the PS with a new synchrotron, optimized for injection at 3.5 GeV and accelerating up to 50 GeV. Such a new accelerator is currently under study [47]. It would solve the problem of ageing PS components and remove the bottleneck at injection in the SPS, preparing for LHC luminosity (and possibly energy) upgrades. 14

25 Chapter 4 Design 4.1 Main design choices The choice of technologies is closely related to the parameter choices in Chapter 2, the technological feasibility of accelerating structures, as outlined in the following sections, and the possibility to re-use equipment from CERN (e.g., LEP RF) or other sources (e.g., IPHI RFQ). While the technology of the superconducting section completely changed from CDR1 to CDR2, the main technological choices for the normal-conducting section (up to 90 MeV) remain valid. Any modern high-intensity linac injecting into a circular machine relies on H charge-exchange injection and implies the use of an H source. Moreover, a low-energy beam chopper is required to reduce capture losses during the ring injection. For the H source development at CERN, it was decided to further develop the existing DESY RF volume source (Section 4.2.1), improving its performance to reach the SPL specifications. After a magnetic LEBT (two solenoids), an RFQ accelerates the beam from 95 kev to 3 MeV. It is foreseen to re-use the RFQ of the French IPHI Collaboration (Section 4.2.2). Originally foreseen for CW operation, this device is under construction and will be tested first in France and then at CERN in the 3 MeV test stand [48] which is currently being prepared. Following the RFQ a beam chopper allows gaps to be created in the bunch train corresponding to the transitions between RF buckets of a subsequent circular machine, reducing the injection losses to tolerable levels. In order to avoid partially filled bunches which can yield losses in the linac itself, the rise-time for the deflecting field has to be less than 2 ns. Furthermore the chopper can also provide gaps for the ring ejection kicker, and/or gaps which allow fractions of the full energy beam to be deflected to different users (see the proposed EURISOL operation scheme in Section 2.2). A performance test of H source, LEBT, IPHI RFQ, and chopper is foreseen partly within the European HIPPI Collaboration [50, 49] and then in a dedicated 3 MeV test stand at CERN [48]. The RF frequency of MHz for the normal-conducting part of the linac is determined by the existing LEP klystrons. From 90 MeV onwards, MHz klystrons are foreseen to power the normalconducting Side Coupled Linac (SCL) up to 180 MeV. From there, MHz superconducting elliptical multi-cell cavities raise the beam energy up to its final value of 3.5 GeV. The higher frequency reduces the size (and cost) of the structures. It allows higher RF gradients to be used in the SCL and it is well suited for bulk-niobium cavities in the superconducting part. After the chopper line, a classic Alvarez Drift Tube Linac (DTL) is used to accelerate the beam to an energy of 40 MeV. Because of the demands of high current and low losses, short focusing periods are imperative in this energy range. Having its maximum shunt impedance at around 20 MeV, a DTL is considered the most suitable structure for this section (see also Section 4.3.1). From 40 MeV onwards, longer focusing periods can be accepted and one can separate the quadrupoles from the RF structure. This measure slightly raises the real estate shunt impedance and, more importantly, simplifies considerably the construction and alignment of the structures. Therefore a Cell-Coupled DTL (CCDTL) was chosen to cover the energy range from 40 MeV to 90 MeV (Section 4.3.2). At higher energies, drift tubes in DTL-type structures become very long and it becomes more efficient to use RF structures with a gap distance of ˇ=2 rather than ˇ as used in DTL-type structures. Switching at the same time to a higher frequency, the structures become small enough, so that they can be machined out of solid copper at a reasonable price. For the SPL baseline design an SCL was chosen to accelerate the beam up to 180 MeV. However, as the construction is still relatively expensive and since the shunt impedance that can be reached is still relatively low (see Section 4.3.3), it is of interest to consider a superconducting alternative for future revisions of the design. Above 180 MeV the technology of superconducting bulkniobium elliptical multi-cell cavities is well established (see Sections 2.3, ), and considered to be 15

26 4 DESIGN the most economical approach to reach high-intensity, high-energy beams. For the SPL it is foreseen to use two families of five-cell elliptical cavities (ˇ D 0:65, and ˇ D 1) to cover the energy range from 180 MeV to 3.5 GeV. This approach will reduce the amount of R&D as well as the production costs for these structures. Transfer lines transport the beam from the linac to the various users (i.e., the PS machine, an accumulator/compressor ring, ISOLDE/EURISOL). At 3.5 GeV a minimum bending radius of 200 m has to be observed for these lines to keep H stripping at an acceptable level (see Section 6.1). An additional design constraint may be considered for the operation as a EURISOL driver: depending on future target studies it may be desirable to extract a fraction of the beam at lower energy. In this case one can use fast kickers at the energy level in question and divert part of the beam to a transport line running in parallel to the main linac. This option has not been studied in full detail but is considered to pose no significant technical challenge. 4.2 Front end H source and low-energy beam transport (LEBT) Today there are two major types of H source: surface and volume sources. A review of their technical principles, advantages and disadvantages is given in Ref. [51]. Not all of these ion sources are well suited for accelerator operation where intensity, short- and long-term stability, reliability, and easy maintenance are of major importance. Reference [52] gives an overview of operational sources and source developments for high-intensity accelerators. The technology of existing sources raises a number of questions. In many cases caesium is used to produce high H currents, but caesium pollution of the beam can give rise to voltage holding problems in the following accelerator structures. And the caesium injection lines can be clogged by caesium hydride, leading to lifetime limitations. The source emittance is defined mainly by the source structure and the principle of the H creation and can only be influenced to a small extent. For a fully optimized source there is a lower limit of the emittance that can only be overcome by cutting the beam intensity at the same time. The lifetime of filaments, cathodes or antennas inside the plasma chamber is a limiting factor for the reliability. The requirements of 80/40 ma H current, 0.57/0.8 ms beam pulse, 50 Hz repetition rate, and 0.25 mm mrad normalized r.m.s. emittance for neutrino/eurisol operation cannot be met by one of the existing operational sources. For the H source development at CERN, it was to decided to start with an existing successful source design like the DESY RF volume source (Fig. 4.1) [53, 54] and then try to improve the design up to the SPL specifications. The DESY source routinely delivers 40 ma of H at an extraction voltage of 35 kv. The present pulse length is 0.1 ms, the repetition rate 5 Hz and the emittance 0.18/0.16 mm mrad (r.m.s. normalized). The source runs without caesium and has a lifetime of up to hours. Currents of 54 ma have already been extracted in experimental set-ups [53]. At the moment the pulse length is limited by the available RF generator. Further tests will show if there is a physical limit to get longer pulses. A collaboration with DESY has to be set up to exchange information and knowledge about the source. Development work of other groups with similar source types (e.g., at SNS) will also supply a valuable input. To meet the SPL duty cycle specifications a major re-design will be necessary (e.g., for the source cooling). Since the extraction voltage of the DESY source is lower than the 95 kv necessary for the injection into the IPHI RFQ, there are two possible ways to adapt the source. The first would be a re-design of the extraction system for 95 kv and the second, and so far more favourable one, is post-acceleration from 35 kv to 95 kv. The source and the LEBT have to be considered together from the early design stage. Twosolenoid LEBT designs as for the CERN RFQ2 injector [55] are well proven and allow a precise matching 16

27 4.2 FRONT END Fig. 4.1: The DESY volume H source [54] Fig. 4.2: LEBT scheme [56] (FC Faraday cup, BT beam transformer) to the RFQ, leaving enough space in the line for beam diagnostics and other equipment. A pre-chopper for shaping the pulse at low energy may also be necessary. A possible design for the Linac4 LEBT based on available equipment (two solenoids and their power supplies) is shown in Fig. 4.2 [56]. The LEBT for the SPL will need new DC magnets and the design has to be based on the final SPL H source and experience with the Linac4 LEBT. With a pressure in the range of 10 5 mbar, space-charge compensation in excess of 90% should be achieved, leading to a transverse emittance growth of only 2%. However, at this pressure, stripping losses of 8% have to be expected. The gas pressure and type can be optimized to maximize the transmission to the RFQ within its acceptance. 17

28 4 DESIGN Radio frequency quadrupole (RFQ) After the ion source and the LEBT, an RFQ will perform the beam capture and bunching and will accelerate the beam from 95 kev to an energy of 3 MeV which is considered as the optimum for beam chopping. The IPHI (Injecteur de Protons de Haute Intensité) RFQ that is foreseen for operation in Linac4 and the SPL is the result of a CEA and IN2P3 joint effort to develop a CW injector [57] for highintensity applications. A collaboration agreement has been signed with CERN, specifying that the RFQ will be transferred to CERN after high-intensity tests with a proton source at CEA Saclay. The RFQ, which operates at MHz, is 6 m long and consists of three segments that are resonantly coupled by means of two coupling cells. Each segment is built with two modules of 1 m length that are mechanically coupled to each other (Fig. 4.3). The end of the third segment, the RFQ output, houses a fringe field cell. Fig. 4.3: One of six modules of the IPHI RFQ accelerating efficiency and voltage (V/100 kv) vane voltage V efficiency A focusing Factor B focusing factor aperture (cm) and vane modulation vane aperture a vane modulation factor m synchronous phase φ synchronous phase (deg) RFQ longitudinal coordinate (cm) RFQ longitudinal coordinate (cm) Fig. 4.4: (a) Vane voltage, acceleration efficiency, and focusing factor; (b) aperture, modulation, and synchronous phase The evolution of the main parameters all along the RFQ is shown in Figs. 4.4 (a) and 4.4 (b). The vane voltage varies between 87 kv and 122 kv in order to keep the surface field at a constant value of 18

29 4.2 FRONT END 31 MV/m (1.7 kilpatrick) all along the structure. The RFQ has been designed for 100 ma beam current and, in order to assure the best beam transmission, the gentle bunching section extends over most of the accelerator length. For the SPL current requirements of 70 ma, 99.5% of the beam is transmitted (see Fig. 4.5). The overall power dissipation within the pulse is depicted in Fig. 4.6, assuming a maximum beam current of 70 ma to be accelerated within the pulse. 100% transmission 96% 92% 88% accelerated transmitted 84% 80% input beam current [ma] Fig. 4.5: Transmitted and accelerated beam in the RFQ as a function of the input beam current kw kw kw 99.0 kw 87.4 kw 95.5 kw kw kw 0.1 kw 4.1 kw 16.1 kw 41.9 kw 72.7 kw MW +20% Cu : MW kw 74.5 kw Power consumption per cell (kw) Maximum power density : Min = 9.1 W/cm² Max = 14.9 W/cm² Total power MW Cavity power MW Beam power (70 ma) RFQ length (m) Fig. 4.6: Power consumption as a function of the RFQ length. The power dissipated in copper (based on SUPER- FISH simulation) and the beam power are detailed at the level of the module. In order to minimize the impact on the beam dynamics, the presence of multi-polar components in the focusing field has been studied with PARMTEQM. The ratio between the dipole component of the 19

30 4 DESIGN field, responsible for the beam defocusing, and the main quadrupole mode in the RFQ cavity has been kept below 4.5%. The length of the RFQ requires that the resonator be split into three resonantly coupled segments. The two coupling cells, placed every two metres, are critical for the beam dynamics, since in this area the electric field is 5 10% lower with respect to the adjacent vanes. The small emittance growth, which can be observed in the RFQ, is located around these coupling cells. Nonetheless, the perturbation of the beam remains localized and the multiparticle simulations performed with LIDOS.RFQ [58] do not show any relevant deterioration of the beam characteristics at the RFQ output. A summary of the main RFQ parameters can be found in Table 4.1. Table 4.1: Main RFQ characteristics Frequency MHz Total length m Electrical length 7 Number of resonating segments 3 Number of coupling cells 2 Number of cells 560 Aperture radius mm Modulation m r mm Vane voltage kv Peak electric field 1.7 kilpatrick Power in copper (SUPERFISH + 20%) 808 kw Beam power 210 kw Total power 1018 kw Low-energy beam chopping The beam chopper has to establish the required micro-pulse structure by deflecting part of the RFQ output beam onto a dedicated beam dump. The structure is located between the RFQ and DTL, both operating at MHz, and the transfer energy of 3 MeV has been chosen as a compromise between the demands of beam dynamics and the chopper amplifier. While lower beam energies reduce the voltage requirements for the chopper amplifier, they increase the space-charge forces and generally yield more emittance growth. The bunches are spaced by 2.84 ns and come in 0.57 ms pulses repeated at 50 Hz, which means that in order to avoid partially chopped bunches the chopper rise-time has to be of the order of 2 ns (assuming a maximum bunch phase length of 45 ı at MHz). The deflector itself consists of two 50 transmission lines facing each other, which are driven by two pulse amplifiers with opposite polarity signals (500 V). The most challenging mode of operation is foreseen when the SPL injects into an accumulator and compressor ring. To avoid losses at injection and assuming a 44 MHz RF frequency in the ring, the chopper has to remove three out of eight bunches, while for Linac4 the chopper has to provide repetition rates of 1 MHz. For operation in EURISOL mode a section of 0:1 ms has to be removed from each linac pulse. In all modes of operation the chopper can also be used to remove the first s of the pulse, which are generally not very stable during the start-up of the source. A 3D technical drawing of the chopper line is shown in Fig The elements of the complete line are (from left to right, downstream) matching section (four quadrupoles plus buncher cavity), beam chopper (two quadrupoles with chopper plates inside), buncher cavity plus quadrupole plus dump (for the chopped beam), matching section (four quadrupoles plus buncher cavity). 20

31 4.2 FRONT END Fig. 4.7: 3D image of the chopper line with support structure The beam dynamics of the entire CERN chopper line is designed to minimize the plate voltage by increasing the deflection via beam optics (see Section 4.4.4, Fig. 4.29). On the other hand, there is an effort to keep the actual chopper structure as short as possible in order to minimize emittance blow-up during long drifts. At present a 3 MeV test stand is under construction at CERN [48], comprising an H source, the IPHI RFQ (see previous section), and the complete chopper line equipped with newly developed diagnostics to test the proposed chopping schemes (see Section 4.5.3). A list of the elements is given in Table 4.2 and the performance requirements are listed in Table 4.3. Chopper structure Short rise-times in the nanosecond range can be obtained by using travelling-wave stripline structures, where the striplines are meander-folded in order to match the speed of the travelling wave to the beam velocity (see Fig. 4.8). In the present design, a section of 1.2 m is foreseen to accommodate two chopper units of 0.5 m, each housing two deflecting plates of 0.4 m length. To minimize the total length of the chopper line, the chopper units are designed to fit into the bore of existing quadrupoles (see Fig. 4.9). The deflecting plates are driven simultaneously in opposite polarity and water cooling is employed to handle heating from beam losses as well as from the deflecting signal. Since the striplines do not completely cover the deflector plates, one has to apply an effective surface coverage factor for the electric field seen 21

32 4 DESIGN Table 4.2: Elements in the chopper line Element Number Length [mm] Value Comment Long quadrupole I G D 0:8 1:2 T/m With chopper plates inside Long quadrupole II G D 1:4 T/m Short quadrupole I 4 56 G D T/m Short quadrupole II G D T/m Chopper plates Distance: 20 mm Dump Conical hypervapotron Buncher cavities V D 100=140 kv P D 17:7=16:3 kw (measured values) Table 4.3: Beam chopper main parameters Beam energy 3 MeV Overall length 3.7 m Number of chopper structures (inside quads) 2 Number of quadrupoles 11 Chopper plate length 400 C 400 mm Chopper plate distance 20 mm Separation chopped/ unchopped beam 15 mm Chopper structure rise-and fall-time (10 90%) < 2 ns Chopper voltage pulse (per plate) 500 V Effective chopper voltage pulse 400 V Remaining voltage for unchopped beam < 2% Max. chopper frequency 44 MHz Min. pulse length 8 ns Max. chopping factor (duty cycle) 40% Repetition rate 1 50 Hz Chopper deflection angle 5.3 mrad by the beam. In our case this factor amounts to 80%, which means that for a deflection voltage of 500 V (i.e., 1 kv in total, or 800 V effective voltage) at a beam energy of 3 MeV one can achieve a total deflection angle of 5.3 mrad for both units. As shown in Fig. 4.8, the travelling-wave structure consists of a double-meander stripline matched to a beam velocity of ˇ D 0:08. The structure is printed on an alumina substrate with 3 mm thickness. The SPL design is related to the SNS chopper structure, which is also a printed meander structure [59] but which uses notched microstriplines with separating ridges in between. While the SNS structure is printed onto glass microfibre PTFE composite material, CERN uses an alumina substrate which is considered to be more stable against ionizing radiation and which has lower vacuum outgassing rates as well as higher heat-transfer coefficients. Two plates have already been constructed [60] using a MoMn base layer (fired at 1500 ı C in an H 2 /N 2 atmosphere) and several layers of other metals added. Their final shape is determined by a chemical etching process. Results of extensive numerical simulations as well as measurements on the rise- and fall-times and the deflecting efficiency have been presented in Ref. [61]. As an alternative to the MoMn process with chemical etching (done at CERN), a different technique is being proposed by industry. The actual base layer is a silk-screen-printed silver alloy, normally 22

33 4.2 FRONT END Fig. 4.8: The alumina ceramic plates with printed meander structure (MoMn + 30 m Ag); mounting holes for screws are at the sides [60] Fig. 4.9: Complete chopper with vacuum tank mounted in a quadrupole on a beam line support [61] used for brazing ceramic metal interfaces. This layer (a few microns thick) is fired at high temperature under vacuum and has an adhesion strength 100 N/mm 2, which is nearly as good as MoMn. Subsequently, the conductor (copper) thickness is increased up to 30 m by electrochemical deposition, followed by a flash of gold to avoid surface degradation due to contact with air. A 3D view of the complete chopper with vacuum tank, water cooling system and feedthroughs is shown in Fig. 4.9 and a summary of the main parameters is given in Table

34 4 DESIGN Chopper pulse amplifier The electric field between the two plates must be established or removed within the time separating two bunches (2 ns). The most limiting constraint of the foreseen beam patterns is the one that requires the removal of three out of eight bunches. In this case we need an 8.5 ns pulse repeated at 44 MHz. To avoid the displacement of the baseline (0 V for the unchopped beam) across the output pulse mean value, the low-frequency cut-off must practically be DC, which is difficult to achieve with available technology. To relax the low-frequency response and make use of a band-pass amplifier, the bunch removal scheme shown in Fig will be used, where three out of eight bunches can be dumped using the superposition of two 22 MHz square waves with 18% duty cycle (8.5 ns pulse), inverse polarity and shifted by a halfperiod.!" Fig. 4.10: Alternate polarity extraction scheme for the chopper pulse amplifier The removal of a long sequence of bunches is possible by simply increasing the duty cycle for each amplifier to 50% and switching from one polarity to the other. The exact number of extracted bunches will then be obtained by adjusting the first and/or last cycle pulse length and/or with a slight change of the square wave frequency. In the above-mentioned conditions, to limit the baseline displacement to 2%, the low-frequency cut-off must be set around 140 khz while the required 2 ns rise-time imposes a high-frequency response extending to above 175 MHz. To achieve such a wide band response and to manipulate the high required power (5 kw), the pulse amplifier will be based on an ultra-broadband hybrid coupler [62] with a theoretically infinitely high-frequency response independent of the electrical length of the transmission line composing it. Because of the freedom in choosing the line length, the circuit allows the use of any amount of ferrite to shift the low-frequency cut-off to the desired point. Like all hybrid couplers it can be used in sum and difference modes. The latter mode will be used to sum the signal of two identical amplifiers, obtain the polarity inversion required by the scheme, and thus achieve DC cancellation. Using the same basic configuration, a two-way hybrid and a four-way hybrid have been designed for signal combination. In both cases the 3 db response goes from below 20 khz to above 500 MHz. The basic amplifier unit shown in Fig is also built around the hybrid coupler but used in sum mode. To limit the number of cascaded stages and to achieve the required bandwidth, MOSFETS with V ds D 150 V and low capacitances (C iss 50 pf, C rss 2 pf, C oss 20 pf) have been selected. The drain circuit high-frequency cut-off is above 300 MHz and the transient time about 1.7 ns. With a DC supply of 100 V the nominal output pulse amplitude of the unit is 150 V on 25 (Fig. 4.12). The gates are driven by commercial drivers requiring TTL levels at the input. To maintain the wide band response of the circuit the DC supply is fed to the MOSFET drains via the hybrid itself from a zero RF point. In this design the ferrite used to increase the transmission line common mode impedance at low frequency is biased by the supply current. This means that on account of the range of operating currents from 0 to 8 A, the low-frequency cut-off sweeps from 40 khz to 120 khz, which is just below the 24

35 4.3 ACCELERATING STRUCTURES Fig. 4.11: Basic chopper pulse-amplifier unit (simplified schematic) 1341 Acqs 01 Aug 05 15:37:26 Pos Wid ns Rise 1.7 ns Fall ns Period ns Fig. 4.12: Basic chopper pulse-amplifier unit output signal (50 V/div) maximum cut-off frequency of 140 khz. Two such units are combined with a two-way hybrid into a 50, 300 V module. Sixteen modules are then used to achieve the specified 500 V (Fig. 4.13) [63]. 4.3 Accelerating structures Drift tube linac (DTL), 3 40 MeV Following the chopper line at 3 MeV, three Alvarez DTL tanks at MHz accelerate the beam up to an energy of 40 MeV. The first tank is powered by one klystron, whereas tanks 2 and 3 each use two klystrons (with one power coupler per klystron) in order to minimize the number of transitions between structures. Beam dynamics simulations with different beam currents show that the chopper line provides enough flexibility to match different currents into a DTL with fixed magnetic gradients. Therefore permanent magnetic quadrupoles (PMQs) were chosen to provide the transverse focusing because they reduce the cost of the focusing system, simplify the assembly of the drift tubes, and raise the shunt impedance by enabling smaller drift tube diameters. A relatively high accelerating gradient has been adopted (3.3 MV/m in the first tank and 3.5 MV/m in tanks 2 and 3) in order to reduce the linac length, benefiting from the availability of the old LEP klystrons to provide the higher RF power. 25

36 4 DESIGN Fig. 4.13: Combining scheme for four 500 V chopper pulse-amplifier units To further reduce the length of the DTL and to ease the RF tuning, the field ramp of the original tank 1 design in CDR1 [3], [64] has been eliminated without affecting the beam dynamics. In order to avoid too strong focusing at the beginning of tank 1, the initial synchronous phase was raised from 44 ı to 30 ı and the transit time factor was reduced by shortening the gaps. Towards the end of tank 1 the synchronous phase reaches its maximum value of 20 ı, which is then kept constant for tanks 2 and 3 as well as for the remaining normal-conducting linac. The value of 20 ı is a compromise between beam stability and acceleration efficiency. The impact on beam stability of choosing 20 ı rather than the more standard 25 ı has been studied with respect to the development of phase and energy jitter [65] and with respect to r.m.s. emittance growth versus different longitudinal input emittances. In both cases no deterioration of the beam quality could be observed [32]. The missing longitudinal focusing between the tanks is compensated for by lowering the synchronous phase in the first and last accelerating gaps in the tanks. Post couplers are used to stabilize the field and a combination of fixed and movable plungers is used to adjust the frequency of the tanks. The main DTL parameters are summarized in Table 4.4. Table 4.4: DTL main parameters tank 1 tank 2 tank 3 Output energy [MeV] RF frequency [MHz] Gradient E 0 [MV/m] Synchronous phase [deg] 30! Lattice FOFODODO or FODO Max. surface field [kilpatrick] Aperture radius [mm] Tank diameter [m] Drift tube diameter [mm] Length [m] Peak RF power [kw] Max. design duty cycle [%] Klystrons Quadrupole length [mm] No. of gaps No. of post couplers

37 4.3 ACCELERATING STRUCTURES A high-power prototype of tank 1, which could eventually be upgraded to be used in the final installation, is being built in Russia under a collaboration funded by the International Science and Technology Center (ISTC). The collaborating institutes are ITEP, Moscow (RF design and PMQ technology) and VNIIEF Sarov (mechanical design and construction). Although the basic design parameters were decided by CERN, the collaborating institutes elaborated the detailed mechanical design and defined technological solutions suited for their production facilities. The drift tubes are mounted on a girder which is decoupled from the tank and are independently supported (Figs. 4.14, 4.15). The drift tubes are connected to the girder by bellows, which allow their precise alignment. This solution, already adopted for the old Linac2 at CERN, complicates the mechanics but avoids the use of O-rings in the drift tube assembly Fig. 4.14: Cross-section of DTL tank manufactured by VNIIEF, Sarov Between the tanks are transition elements which are under vacuum and which allow space for diagnostics. An elaborate adjustment unit (Fig. 4.16), developed by VNIIEF, allows the precise alignment of the drift tubes with respect to the girder. After the alignment the drift tube is fixed in position with a solid filling and the adjustment unit can be used for the subsequent drift tube. Should a re-alignment become necessary, the filling can be mechanically removed and replaced by a new one Cell-coupled drift tube linac (CCDTL), MeV A layout based on short 352 MHz DTL-like tanks containing two drift tubes (three accelerating gaps) has been selected for the energy range between 40 MeV and 90 MeV. The frequency of this section is kept at 352 MHz in order to use the LEP klystrons and to make possible a simple mechanical construction of the small tanks based on the same technology as the DTL. The tanks are made of copper-plated, stainless- 27

38 4 DESIGN area of bearing locations of post coupler area of bearing area of bearing Fig. 4.15: Side view of DTL tank manufactured by VNIIEF, Sarov Fig. 4.16: Adjustment unit for the alignment of the drift tubes (left: drawing, right: photo) steel elements connected by spring-loaded RF and vacuum joints. In order to avoid a complex power distribution system from the 1 MW klystron to the tanks and to allow simple setting of tank phase and amplitude, the DTL tanks are resonantly coupled by off-axis coupling cavities, leaving sufficient space for quadrupoles between the tanks. The resulting structure, shown in Fig and called Cell-Coupled Drift Tube Linac (CCDTL), is essentially a low-frequency symmetric version of a structure developed at LANL for the APT project [66]. For the chosen inter-tank distance, the resulting focusing period is 7ˇ, short enough to provide acceptable phase advance per period at the starting energy of 40 MeV. A string of three tanks forms one RF module, which is fed by a 1 MW klystron. The resulting module is a chain of five coupled resonators (three tanks and two coupling cells) operating in /2 mode. Under normal operating conditions there is no field in the coupling cells. RF power is fed into the central DTL tank by means of an iris coupling to the side of a waveguide short-circuited at /4 from the iris. The overall layout of the CCDTL is shown in Table 4.5, and Table 4.6 summarizes the main parameters of the CCDTL section. The gradient in each of the eight modules is adjusted in order to divide the RF power evenly between the klystrons, at a power per klystron of 800 kw. As a consequence, the accelerating gradient decreases from 3.9 MV/m to 2.8 MV/m, and the peak surface field decreases correspondingly from 1.6 to 1.3 times the Kilpatrick limit. Figure 4.18 shows the computed shunt impedance of the CCDTL section, compared to that of the DTL and SCL. The transition to the CCDTL allows a partial compensation of the characteristic decrease in shunt impedance of DTL structures. 28

39 4.3 ACCELERATING STRUCTURES Table 4.5: CCDTL layout (three gaps per tank, three tanks per module) Module E 0 P klystron Energy [MV/m] [kw] [MeV] Table 4.6: CCDTL main parameters Input energy 40 MeV Output energy 91.7 MeV RF frequency MHz Gradient E MV/m Synchronous phase 20 deg Number of tanks 24 Lattice FODO Max. surface field kilpatrick Aperture radius 14 mm Length 25.2 m Peak RF power 6.4 MW Max. design duty cycle 14% Klystrons 8 No. of gaps 72 DTL tank coupling cell quadrupole quadrupole βλ 3/2 βλ drift tubes Fig. 4.17: Outline of the CERN cell-coupled DTL, showing electric field lines in one three-cavity module 60 DTL Tank 2 DTL Tank 3 40 DTL Tank 1 CCDTL SCL ZT 2 (MΩ/m) energy (MeV) Fig. 4.18: Computed shunt impedance (linac definition) of DTL, CCDTL, and SCL 29

40 4 DESIGN The development of this particular CCDTL structure took place at CERN during the last four years [35, 67]. After detailed beam dynamics and 3D RF simulations, a high-power prototype made of two half-tanks and one coupling cell was designed and built (Fig. 4.19). It has validated the construction technology, the resonant coupling principle, and the RF coupler design. High-power RF tests foreseen in the SM18 test stand will allow testing of the cooling circuits and will define the RF conditioning procedure. In the frame of an ISTC-funded collaboration, BINP, Novosibirsk and VNIIEF, Snezinsk are currently constructing a complete CCDTL prototype module, whose design is shown in Fig Fig. 4.19: The CERN CCDTL prototype including power coupler at the bottom during low-level RF tests Fig. 4.20: The CCDTL full-module prototype under construction in Russia 30

41 4.3 ACCELERATING STRUCTURES Side-coupled linac (SCL), MeV From an energy of 100 MeV onward, the shunt impedance of DTL-like structures decreases drastically. To maintain a reasonable RF power efficiency, it is convenient to replace the DTL with a -mode structure (i.e., with 180 ı phase difference between the fields in two subsequent gaps) at a higher harmonic of the basic frequency. For the SPL a standard Side-Coupled Linac (SCL) has been chosen at twice the basic frequency (704 MHz) starting from a transition energy of 90 MeV [68]. This particular type of -mode structure, developed at LANL in the 1960s, has been used for many linac projects worldwide. A prototype SCL at higher frequency has already been built at CERN and tested with beam in the frame of the TERA project. The SCL is made of sequences of accelerating cells, coupled via slots to coupling cells placed alternately on both sides of the accelerating cells. Eleven cells are grouped in tanks, spaced by focusing quadrupoles (Fig. 4.21). Multiple tanks can be fed by a single klystron, and are coupled together by bridge couplers, additional short chains of cells joining tanks across the quadrupoles. The coupling cells are needed to resonantly couple the single accelerating cells, transforming the operating mode into the stable /2 mode, which ensures a uniform longitudinal field distribution even in the presence of mechanical and tuning imperfections [69]. coupling cells bridge coupler 1.5 βλ quadrupole Fig. 4.21: Side-coupled linac (SCL) structure for Linac4/SPL The increase in frequency with respect to the previous sections provides higher shunt impedance, and allows the dimensions of the structure to be reduced. However, the larger power dissipation per unit surface requires a different construction technology than for DTL structures. The SCL cavities of the SPL are made of stacks of brazed copper cells. For beam dynamics stability, the same type of focusing (FODO) as in the preceding CCDTL is maintained throughout the SCL lattice, with single quadrupoles between tanks. The aperture radius is 16 mm, about eight times the r.m.s. beam size. The slots between cells provide a cell-to-cell coupling of 3%, sufficient for the required field stability, and it is foreseen to re-machine the slots individually to minimize the effect of machining errors on the coupling factor. The distance between tanks is fixed to 1.5 ˇ to allow sufficient space for quadrupole, bellow, flanges, and some diagnostics. Tanks will be connected by three-cell bridge couplers. Klystrons delivering 4 MW are foreseen, each feeding a module of five tanks. The accelerating gradient has been fixed to E 0 D 4 MV/m, providing a good compromise between length of the structure and RF power consumption, while keeping the peak surface fields below 1.6 kilpatrick. With this gradient, only five klystrons are needed to cover the energy range between 90 MeV and 180 MeV, the last module being made of four tanks. The SCL layout is presented in Table 4.7, and the main parameters in Table 4.8. The development of the SCL is being carried out inside the HIPPI Joint Research Activity, cofunded by the European Union. The RF design is being jointly developed between CERN and the IN2P3 31

42 4 DESIGN Table 4.7: SCL layout (11 cells per tank) Module E 0 P klystron Energy No. of tanks [MV/m] [MW] [MeV] per module Table 4.8: SCL main parameters Input energy 92 MeV Output energy 180 MeV RF frequency MHz Number of tanks 24 Gradient E 0 4 MV/m Synchronous phase 20 deg Lattice FODO Max. surface field kilpatrick Aperture radius 16 mm Length 34.4 m Peak RF power 15.1 MW Max. design duty cycle 14% Klystrons 5 No. of gaps 264 LPSC Laboratory in Grenoble, while the analysis of the behaviour of long SCL chains is being made by INFN Naples. The thermal analysis of the SCL cells is done in Grenoble. A brazed technological copper model, intended for vacuum and low-level RF tests, is going to be built by BINP Novosibirsk as a part of a project funded by ISTC Superconducting linac Since the development of the LEP superconducting cavities, involving niobium sputtered onto copper, there has been considerable progress in the technology of bulk-niobium SC cavities (compare also Section 2.3). In the SPL, two families of bulk-niobium elliptical cavities with a geometrical ˇ of 0.65 and 1.0 have been chosen to cover the energy range from 180 MeV to 3.5 GeV using a frequency of MHz (compare Section ). In the following we review recent progress in this technology and we motivate the chosen accelerating gradients of 19 MV/m and 25 MV/m for the ˇ D 0:65 and ˇ D 1:0 sections. Since the cryo-module design has an important influence on the real-estate (accelerating) gradient, the decision was taken to follow mainly the Tesla Test Facility (TTF) approach, which minimizes the number of cold/warm transitions and uses cold quadrupoles within the cryo-modules. The motivation for this choice and the preliminary parameters are presented in Section Cavity technology The choice of bulk-niobium, multi-cell, superconducting elliptical cavities operating at a temperature of 2 K implies the use of the following technologies [20]: Proper grade material (niobium sheets with Residual Resistance Ratio (RRR) values in the range ). Proper treatment techniques: thermal treatments for stress annealing and hydrogen degassing at ı C; deep chemical etching with Buffered Chemical Polishing (BCP) techniques for removing m of material from the surface; High Pressure Rinsing (HPR) with ultra-pure water to remove any contaminants from the surface. Proper handling procedures (mainly cleanliness, meaning that all work around a cavity and its ancillaries should be performed in a clean room). 32

43 4.3 ACCELERATING STRUCTURES The performance limitations of superconducting cavities are basically caused either by field emission or by the intrinsic limitations of the superconductor material. In the first case the surface electric field on the cavity walls is responsible for the electron emission starting from surface contaminants or protrusions, while in the latter the effect is dominated by the magnetic field level at the cavity surface. Elliptical cavities designed for particle velocities smaller than the speed of light (medium-ˇ cavities) have, for the same frequency, smaller volumes with respect to cavities designed for relativistic particles (ˇ D 1). This is mainly because the cell length of a -mode structure has an active length given by ˇ=2, while the cell diameter does not depend, to first order, on the design ˇ, but only on the frequency. In other words, the ratio of peak electric (or peak magnetic) field to the cavity accelerating field grows with decreasing design ˇ. The geometrical shape of the superconducting medium-ˇ structures can be optimized for the application needs by using a proper parametrization, which allows independent tuning of electromagnetic and mechanical characteristics [70], such as peak field ratios, cell-to-cell coupling, R/Q, Lorentz-force detuning coefficients, and mechanical stresses under operating conditions. Several bulk-niobium, reduced-ˇ, multi-cell cavities at different frequencies have been designed, fabricated and tested in the last decade, for various high-power proton linac applications. The tests of these structures yielded results which are consistent, in terms of peak fields, with those of the extensive experience gained with the TTF production and cavity treatments. The measured electromagnetic characteristics of the cavities listed in Table 4.9 are the basis for the estimated performance of newly developed SPL cavities. Table 4.9: Medium-ˇ multi-cell cavity parameters (bulk-niobium structures only) Project f ˇ E peak =E acc H peak =E acc N cell (R/Q)/N cell Ref. [MHz] [mt/(mv/m)] [] RIA [26] TRASCO [22] SNS medium-ˇ [25] CEA/CNRS [24] SNS high-ˇ [25] TTF [71] (R/Q) uses the linac definition. All the cavities in Table 4.9 have been fabricated, handled, and tested using the procedures briefly outlined in the previous paragraph. Single-cell prototypes, which can be found in the literature, have not been considered. Single-cell cavities have much lower values of E peak =E acc because of the electric field penetration in the large bore of the end tubes, and they are less sensitive to field emission effects during tests. In addition, the chemical and rinsing procedures are more effective for these simpler geometries. As a consequence, single-cell structures generally reach higher accelerating fields than the corresponding multi-cell structures [72]. Figure 4.22 shows the design peak (electric and magnetic) field ratios (with respect to the accelerating field) for the cavities listed in Table 4.9, as a function of the cavity design ˇ. It illustrates the increase of the E peak =E acc and H peak =E acc factors with the decrease of the design ˇ of the structure. Figure 4.23 shows the behaviour of the geometrical R/Q value per cell of the cavities in Table 4.9. We have used the linac definition of the shunt impedance definition [R D. V c / 2 =P c ], where V c is the voltage provided by the cavity (at the nominal ˇ) and P c the power dissipated on the walls. Again, the geometrical R/Q value per cell decreases almost linearly with ˇ. 33

44 4 DESIGN Fig. 4.22: Peak field ratios for the cavities listed in Table 4.9 Fig. 4.23: Geometrical factor R/Q (linac definition) per cell of the cavities in Table 4.9 It is not straightforward to make a complete comparison of the TTF experience (which started 15 years ago and produced industrially in four production batches a total of more than 100 multi-cell cavities, performing hundreds of CW vertical tests at the DESY facilities) with the results of the very few cavity prototypes built by the other programmes listed above, except for SNS, for which a comparable number of resonators have been built and tested [73]. An analysis of the TTF cavity data [74] shows a definite learning curve effect on the achieved cavity performances during the first three production batches [75]. As the performance goal of the TTF 34

45 4.3 ACCELERATING STRUCTURES Collaboration was shifted to higher accelerating gradients (in the range of MV/m) in view of the TESLA Linear Collider scale (and now of the global design effort for the International Linear Collider [76]), the standard chemical treatment of the TTF cavities was switched to etching by electrolytical polishing (EP), a surface preparation method which yields smoother surfaces with respect to the BCP procedure and is more promising for achieving performances closer to the material limitations in terms of surface magnetic fields [77]. Nonetheless, the RF performances of the BCP-treated cavities of the third production batch of the TTF cavities achieved a maximum accelerating field of 28.3 MV/m, with an r.m.s. spread of 1.3 MV/m and an average field emission onset greater than 22 MV/m. In the following comparison with the performance of other cavities, the TTF cavities are accounted for in an operation extending between 22 MV/m and 28 MV/m average maximal accelerating field. It should be noted, however, that chemically etched cavities already reached accelerating fields in excess of 30 MV/m, and electro-polished cavities were capable of reaching even higher fields, up to nearly 40 MV/m in the absence of field emission (only considering BCP chemical etching). Concerning SNS, the performance data on the cavity prototypes was described in Ref. [25] (with 17 MV/m and 20 MV/m of E acc for ˇ D 0:61 and 0.81, respectively). More recent data, with production statistics, has been presented in Ref. [73] and is summarized in Table The main limiting factor Table 4.10: SNS cavity performances Cavity Eacc max [MV/m] Field-emission onset [MV/m] Number tested ˇ = ˇ = reported from the SNS production experience, and the main cause of the relatively high rejection rate experienced in the production phase, is an early field-emission onset (the SNS acceptance value asks for a peak field of 27.7 MV/m without field emission). Other medium-ˇ elliptical structures for which RF test data is available and is included in Table 4.9 are the TRASCO five-cell cavities (two ˇ D 0:47 structures [22]), the RIA six-cell cavities (three ˇ D 0:47 structures [26]), and the CEA/CNRS five-cell ˇ D 0:65 cavity [24]. Figure 4.24 shows their surface electric field corresponding to the maximum gradients and field-emission onset. It is worth noting that, in terms of peak surface electric field, the prototypical medium-ˇ proton cavities considered show comparable performances to the average TTF cavity production of the third batch. The standard TTF cavity preparation procedure (chemical etching followed by high-pressure rinsing stages with ultra-pure water) seems to be suitable for the reduced-ˇ geometries in this range (down to the extremely compressed 0.47 structures). Moreover, no indications for any dependence of the peak surface field on the cavity frequency seem to be evident from the data. Figure 4.25 shows the data for the same set of cavities in terms of peak surface magnetic field. Owing to the stronger dependence of E peak =E acc as a function of ˇ with respect to the ratio H peak =E acc (see Fig. 4.22), the electric field limitations are more rapidly reached in the lowest ˇ structures, and the larger spread seen in the data for the ˇ D 0:47 structures does not come from magnetic field limitations, but from electric field limitations. The fact that the lower-ˇ structures are more sensitive to electric field effects than to magnetic ones is explained by the behaviour of the ratio H peak =E peak, shown in Fig as a function of the structure ˇ. As seen from the figure, the lowest-beta structures tend to be penalized by the drastic reduction of the cavity capacitive volume, which implies that generally the field emission limitation is reached well before the magnetic limitation. 35

46 4 DESIGN Fig. 4.24: Peak surface electric field (MV/m) corresponding to the maximum gradients reached in the cavity tests (down triangle) and to the onset of field emission, FE (up triangle) The analysis presented here takes into account the wide experience accumulated in the last 15 years by the TESLA Collaboration on the reliable production and operation of high-gradient multicell structures, the ongoing work performed both for the XFEL Project and the ILC Collaboration for reaching even higher performance goals, and the learning curve observed during the TTF production. Considering these experimental results, we base the design of the superconducting SPL section on the assumed parameter values given in Table Table 4.11: SPL superconducting linac design parameters Maximum peak surface electric field 50 MV/m Maximum peak surface magnetic field 100 mt Cavity quality factor at 2 K Accelerating gradient (ˇ D 0:65) 19 MV/m Accelerating gradient (ˇ D 1:0) 25 MV/m R/Q (ˇ D 0:65) 290 R/Q (ˇ D 1:0) 570 Frequency MHz Number of cells 5 We also conclude that the maximum peak fields show only little dependence on the operating frequency, meaning that there is no motivation to introduce a third frequency step (e.g., to 4352:2 MHz) in the SPL. 36

47 4.3 ACCELERATING STRUCTURES Fig. 4.25: Peak surface magnetic field (mt) corresponding to the maximum gradients reached in the cavity tests Fig. 4.26: Ratio of peak magnetic to peak electric fields 37

48 4 DESIGN Criteria for cryo-module design The real-estate gradient of a 2 K superconducting linac does not depend uniquely on the RF structure accelerating gradient, but more importantly on the technological choices that determine the layout of the cryo-modules. Each cold-to-warm transition at the beam-line level requires a certain length and reduces the overall filling factor. Currently, there exist two proven designs, with different relative merits, for a GeV-sized 2 K superconducting RF accelerator: the TTF/TESLA/ILC [78] and the SNS cryo-modules [79]. The TTF module has been conceived as the building block for a TeV-sized superconducting electron linac. Thus, the design aims for very low static losses (due to the huge size of the accelerator with superconducting RF structures), high filling factors (to minimize the linac length), and low cost. The module contains eight cavities, a superconducting magnet package for beam focusing every three modules, and integrates the distribution lines of the cryogenic fluids. Tens of modules are connected together to form independent strings from the vacuum and cryogenic point of view, limiting the number of cold-to-warm transitions along the beam line. These choices imply that the substitution or opening of a single module leads to a warm-up/cool-down cycle for the whole string, and hence high reliability needs to be guaranteed in the cold mass. The TTF concept is still the baseline for the ongoing ILC cryo-module design [80]. The SNS module, on the other hand, hosts few (3 4 in the two beta sections) cavities, does not contain any focusing elements the quadrupole doublets are located in the warm sections between modules and has been designed to be quickly disconnected from the beam line (in less than a day a SNS linac module can be exchanged). These design criteria, together with the moderate energy of the SNS superconducting linac (extending from 200 MeV to 1 GeV), led to a much lower filling factor than the TTF case. For the 3.5 GeV SPL linac at the CERN site, filling factor considerations play an important role, and this strongly motivated the choice to base the linac design on long interconnected modules with several (6 8) cavities and cold focusing magnets in each unit. By using such a layout, the 180 MeV 3.5 GeV energy range can be covered with the two ˇ D 0:65 and ˇ D 1 sections in approximately 350 m. At the high duty cycle envisaged for the SPL linac, the main heat loss contribution at the 2 K level is given by the RF dynamic losses on the cavity walls. As in the TTF module, a K thermal radiation shield, with Multi-Layer thermal Insulation (MLI), screens the inner mass from the 300 K environment of the vacuum vessel and provides thermalization of penetrations (i.e., couplers, supports and cabling) at high temperatures. A second inner shield, operating at 5 8 K with MLI insulation, further shields the cold mass from the thermal radiation of the outer shield and provides a further possibility of thermalization of the penetrations. A preliminary estimation for the main heat loads on the cryogenic circuits in the two modules is given in Table The static loads reported here are scaled from the measured values in the TTF modules [19], increasing the thermal radiation contribution (amounting to approximately 1/3 of the total) in the higher temperature circuits to take into account the larger cross-section needed to accommodate the larger 704 MHz cavities. The dynamic RF load at a 6% duty cycle has been calculated by using the cavity parameters reported in Table 4.11, according to the extrapolation given in Fig A dynamic load at 2 K accounting for a beam loss level of 1 W/m has been included. The heat load summarized in Table 4.12 represents a preliminary estimation and contains only a rough estimation of the dynamic load contribution from couplers and Higher Order Mode (HOM) absorbers, for which a detailed design, and corresponding calculations, are still missing. The results from the SNS linac commissioning will provide experimental data allowing these estimations to be refined for the high duty cycle case of the SPL. Considerations on the cryogenic system architecture needed for the superconducting section of the SPL linac are expressed in Section

49 4.3 ACCELERATING STRUCTURES Table 4.12: First estimations on heat losses in the SPL modules assuming a 6% duty cycle ˇ D 0:65 module (six cavities, m length) Heat load 2 K [W/m] 5 8 K [W/m] K [W/m] Static load Dynamic RF load 1.9 HOM Coupler} 0.06 Beam losses 1 ˇ D 1 module (eight cavities, m length) Heat load 2 K [W/m] 5 8 K [W/m] K [W/m] Static load Dynamic RF load 4.2 HOM Coupler} 0.13 Beam losses 1 Estimated to be 40% of the dynamic RF load (see Section 4.7.2). } Estimated to be 3% of the dynamic RF load (see Section 4.7.2) Layout of the superconducting section The transition energy between normal and superconducting structures was chosen to be 180 MeV. Above this energy the real-estate gradient of currently available superconducting elliptical cavities is substantially higher than for normal-conducting structures. Below this energy, we consider that there is still need for R&D before superconducting structures can be used at low cost in a pulsed high-power proton linac. The different cavity types and transition energies are derived from the optimization of length and cost. To minimize development costs, complexity of maintenance and stock of spare parts, only two different cavity shapes will be used, one for the medium-energy range with a geometrical ˇ of 0.65 and one for the high-energy range with ˇ D 1. The operating frequency is 704 MHz for both structures and the gradients correspond to the values in Table The optimization process considers several combinations of different architectures taking into account cavity length, transitions between cavities, number of cells, and achievable gradients. The longitudinal phase advance at zero current per lattice must be lower than the transverse one to avoid emittance exchange between the planes due to space-charge resonances. The phase advance per metre is kept smooth along the superconducting linac in order to achieve a current-independent structure. This condition also simplifies the beam matching at transitions. The setting of the synchronous phase ensures the continuity with the preceding normal-conducting section and provides sufficient acceptance for the injection into the superconducting linac. The synchronous phase varies linearly from 20 ı to 15 ı with the final value being reached at 50 m. It has been estimated that 15 ı provides enough stability to manage static and dynamic errors of the RF system. The optimization procedure converges to the architecture given in Table Quadrupoles are included in the cryo-modules. For the medium-energy section one cryo-module contains two focusing periods, whereas in the high-energy part one period is housed in one cryo-module. In addition to the reduced cryogenic losses, the reduction of cold-to-warm transitions minimizes the length of the linac. This choice, coupled with the high gradients, yields a linac length of 342 m. The maximum power per coupler is 543 kw for the first section and MW for the second section [Fig. 4.27(a)]. Symmetric five-cell cavities maintain field flatness over a wide tuning range in both sections by providing an identical deformation of both end-cells. Their large apertures minimize the risk of inadequate damping of higher order dipole modes which could be excited by the beam. Stiffening 39

50 4 DESIGN Table 4.13: Layout of the superconducting sections assuming five cells per cavity Sections W in W out Cavities per No. of No. of Cryo- Aperture Length [MeV] [MeV] module cav. klystr. modules radius [mm] [m] ˇ D 0: ˇ D 1: Total Used for loss calculations. rings are welded between cells in order to minimize the Lorentz-force detuning. The main characteristics of the medium- and high-ˇ cavities are listed in Table 4.11 and the evolution of the real-estate gradient is shown in Fig. 4.27(b). c avity po wer ( kw ) TraceWin - CEA/DSM/DAPNIA/SACM p o sitio n ( m ) Fig. 4.27: (a) Power per coupler/cavity in the SC linac e ne rg y g ain pe r m e te r ( MeV/m ) Trace Win - CEA/DSM/DAPNIA/SACM p o sitio n ( m ) (b) Real-estate electric gradient in the SC linac Be a m Lin a c The dimensions of the cryo-modules in Table 4.14 include all cold components [cavities, helium vessel, tuner, superconducting quadrupoles, Beam Position Monitors (BPMs), steerers, etc.]. The superconducting quadrupoles operate with integrated magnetic field gradients lower than 4 T for an aperture radius of 100 mm and a length of 450 mm (see Fig. 4.28). Table 4.14: Lengths in the superconducting linac (m) Section Cavity Quadrupoles Cryo-module Between cryo-modules Period ˇ D 0: ˇ D 1: Beam dynamics General approach Beam dynamics considerations have influenced the choice of the structure parameters from the first conception stage onwards. The main guidelines have been the control of losses and the minimization of r.m.s. emittance growth ( " r ) and halo development. In order to implement these guidelines, much care has been taken to keep the following constraints: a) a zero-current phase advance always below 90 ı, for stability; b) a longitudinal-to-transverse phase advance ratio (with current) between 0.5 and 0.8 in order to avoid resonances; and c) a smooth variation of the transverse and longitudinal phase advance per metre. 40

51 4.4 BEAM DYNAMICS TraceWin - CEA/DSM/DAPNIA/SACM Inte g ral o f g radie nt ( T ) P o sitio n ( m ) Fig. 4.28: Integrated magnetic gradient for all SC quadrupoles The accelerator optimization has been conducted in sections while an end-to-end simulation has validated the choices and has allowed transitions to be smoothed wherever needed. Particular attention has been paid to the matching at each transition point: an initial envelope-type matching has been subsequently refined by a multi-particle type matching on a cloud of particles generated at the low-energy end. The starting point of the simulations is the Low-Energy Beam Transport (LEBT) between the source and the first stage of RF acceleration. In this section the space-charge electric field of the quasi- DC beams is partly compensated for by the trapping of positive ions. These are created by ionization of residual gas molecules, and the rapid expulsion of the free electrons by the electric field of the beam. This partial beam neutralization influences the emittance growth and halo development, and it can significantly change the beam distribution at the start of the acceleration. When simulations from the source were not available, two different distributions were tracked through the accelerator: a 4D waterbag and a Gaussian. The different evolution of these two distributions is expected to give a boundary to the behaviour of the real beam. Most of the simulations have been computed with macro-particles, giving therefore an accuracy for any global quantity reported in this section of about 0.5%. The figures of merit for the beam dynamics are the losses, the r.m.s. emittance growth, and the sensitivity to errors. In the following a description of the beam dynamics in each linac section is given. Unless otherwise indicated, matching and beam transport are performed with the code packages TraceWin/Partran [81] using 3D space-charge routines. The quadrupoles are simulated with a hard-edge linear formalism. The fields in the cavities are computed via a sophisticated multi-gap model to take into account the effect of a finite bore radius. The radial dependence of the field is extrapolated with a Bessel function. Most of the simulations have been cross-checked with other codes LEBT (95 kev) The LEBT must ensure the matching of the beam from the source to the RFQ, while minimizing the emittance growth. The choice has been made to employ magnetic lenses rather than electrostatic ones for the following reasons: a) to avoid the high-tension discharge problems of electrostatic lenses; b) the possibility of space-charge compensation; c) the positive experience with magnetic lenses in the Linac2 LEBT; and d) the availability of spare pulsed solenoid lenses from Linac2. Simulations of the beam transport in the LEBT have been peformed with the multi-particle code PATH [82], using calculated field maps of solenoid magnets. Beam input conditions have been based on measurements of the H C ECR SILHI source at CEA Saclay, which had an increased emittance of 41

52 4 DESIGN 0.25 mm mrad (normalized, r.m.s.). The beam is created with uniform density in the transverse plane, and the space-charge forces are calculated using the SCHEFF routine, imposing a circular symmetry of the space-charge field. Calculations use a linear compensation of the beam intensity to simulate the effect of space-charge neutralization. Several groups have reported successful use of this compensation model to explain measurements of beam transport lines. However, understanding and limiting the creation of beam halo will require the understanding of the non-linear aspects of the compensation process [83]. These simulations have yielded the following results: 1. The nominal beam (" r D 0:25 mm mrad, 70 ma) can be transported and matched to the RFQ with no losses. 2. With full space-charge, emittance growth can be limited to 25% by limiting the dimension in the solenoids to avoid lens aberrations, which in turn would lead to a strongly non-linear space-charge field. Such a requirement leads to a short distance between the source and the first solenoid. 3. With 90% space-charge compensation, the emittance growth can be limited to 2%. 4. There is negligible emittance growth due to misalignments of the beam or solenoids, up to 2 mm and 20 mrad. It is required that the beam be steered into the acceptance of the RFQ. From the simulations it is evident that beam neutralization plays a very important role in the LEBT. Calculation and measurement of the beam potential, ion oscillations, and electron oscillations by Soloshenko [84] suggest working at a pressure close to or slightly below the critical pressure (corresponding to the pressure at which the beam potential is zero). As this critical pressure is estimated to be in the region of mbar, it is proposed that a gas injection system into the LEBT be foreseen to allow the pressure to be varied independently. The time () for the production of sufficient ions through beam impact ionization is given by D 1=nv, where n,, and v are the residual gas density, the beam-molecule ionization cross-section, and the beam velocity, respectively. Typical cross-sections for H 2 ionization by 100 kev protons are cm 2 [85], leading to a compensation time of approximately 8 s at mbar. This relatively high pressure will lead to gas stripping of the H beam. The measured cross-sections for electron detachment of H to H 0 or H C (see Refs. [86], [87]) are shown in Table 4.15 for some gases for 100 kev H ions. Table 4.15: H single gas stripping cross-section (in cm 2 ) for selected gas types, for ion energies of 100 kev H 2 [86] He [86] N 2 [86] Ar [86] Ne [87] Kr [87] Xe [87] 1; Value from interpolation of data at 50 kev and 150 kev in Ref. [86]. The mean free path is then of the order of D kt =P D 8:1 m for a cross-section of cm 2, and a pressure of mbar at T D 293 K, with k being the Boltzmann constant. A 1.42 m long LEBT would then have stripping losses of >15%. If the pressure required for good compensation is of the order of mbar, the losses would be reduced to < 4%, and lower still if low cross-section gases are used. An emittance measurement at the end of the LEBT could indicate the optimum pressure settings. Data also exist in the same references for the double electron detachment cross-sections from H to H C, but these are typically an order of magnitude smaller at these low energies, and hence are ignored. However, the double detachment from H to H C will produce a low-intensity beam of protons. If the detachment occurs outside a solenoid field, the protons will be focused into the RFQ (unless there is a strong steering force). The effect of such a beam composing less than 1% of the H current should not be ignored for the SPL. 42

53 4.4 BEAM DYNAMICS RFQ (95 kev to 3 MeV) The RFQ used in the SPL set-up is the RFQ of the IPHI project, which has been extensively documented in Refs. [57], [88], and [89]. The main feature of this RFQ is the capability of accepting a CW beam with varying beam current up to 100 ma. In our particular case, the reference peak beam current is 70 ma and the input r.m.s. normalized emittance is 0.25 mm mrad Chopper line and matching to the DTL (3 MeV) After the first stage of acceleration it is necessary to introduce a section capable of removing a few of the micro-bunches in the pulses. This operation is done at low energy with the aim of removing microbunches that would eventually be lost at the injection in a circular machine with a typical operating frequency of the order of some tens of megahertz. This operation is done by a fast-switching electrostatic field provided by the chopper (see Section 4.2.3). On account of the conflicting needs of high voltage and fast rise-time, the chopper is a relatively long object compared to the focusing periods in the RFQ and DTL. Housing it in the beam line implies an abrupt modification of the phase advance per metre. In order to disrupt as little as possible the continuity of the phase advance, the chopper is integrated in a FODO period with a length of 1.6 m. A total of four FODO periods assure the matching from the RFQ to the chopper and from the chopper to the DTL. The FODO period housing the chopper is also essential for the chopping action, as it amplifies the effective kick of the chopper (5.3 mrad kick for 400 V effective voltage) by adjusting the appropriate phase advance for the beam centre (almost 90 ı from the centre of the chopper to the middle of the dump). With 11 quadrupoles, already existing at CERN, and three new buncher cavities in the chopper line, the beam can be chopped with only 0.3% of the particles remaining in the empty bunches and with an emittance increase of < 20% in the three planes of the unchopped beam. The losses of the unchopped beam take place partly on the chopper plates (1%) and partly at the dump (6%). It should be noted that the dump can be used also as a rudimentary collimator to limit the emittance growth to a few per cent at the expense of increased beam loss at the dump (12%). The envelope of the chopped and unchopped beam is shown in Fig Fig. 4.29: Beam envelope in the chopper line: chopped beam (top) and unchopped beam (bottom) 43

54 4 DESIGN DTL, CCDTL and SCL (3 MeV to 180 MeV) The dynamics and the structure parameters of the acceleration section between 3 MeV and 180 MeV have been optimized as a whole, although this part is composed of three different RF structures. We have designed the three structures globally, taking into account that a variation of parameters in one of them must be followed by the others. The longitudinal phase is ramped up in the first two metres of the DTL from 30 ı to 20 ı and then kept constant until the end of the SCL. At the transition between tanks and/or structures where one or more gaps are missing, the phase of the neighbouring cavities is adjusted to cope with the transition and to ensure a smooth variation of the longitudinal phase advance. The transverse confinement and matching is provided by a system of quadrupoles, using an FFDD configuration in the first DTL tank (2.56 m) and an FD configuration afterwards (> 10 MeV). The period length changes from 2 ˇ at low energy to 11 ˇ at 90 MeV. The reason for starting with an FFDD configuration is purely technical and is due to the difficulties of achieving a strong enough integrated gradient in the limited space available for the quadrupole in the first drift tube (45 mm mechanical length available for a quadrupole). A scheme with FD configuration everywhere has been studied and it gives some advantages with respect to sensitivity to beam errors [90]. However, since it is not the baseline scheme, it will not be presented in this report. The transverse and longitudinal (phase) r.m.s. envelopes for a matched Gaussian beam generated at the entrance of the DTL are shown in Fig The transitions between focusing structures preserve the continuity of the phase advance per metre. r.m.s. x size (mm) length (m) r. m.s. phase (deg) length (m) Fig. 4.30: Transverse, x (top) and phase (bottom) r.m.s. envelopes along DTL, CCDTL, and SCL. The dotted lines mark the transition between structures. At all times during the acceleration (except at one point), the ratio of transverse to longitudinal phase advance is chosen to respect Hoffmann s stability criterion [91]; hence the emittance exchange during acceleration is under control. Table 4.16 summarizes the r.m.s. emittance growth in each section for a Gaussian and a uniform input beam distribution without errors Superconducting section (180 MeV to 3.5 GeV) To minimize r.m.s. emittance growth and halo formation, several techniques have to be combined to compute the parameters of the channel. For this purpose, the same principles as introduced in the normalconducting part are applied for the superconducting section: a) the phase advance per metre (average beam size) has to be kept as smooth as possible to maintain the equilibrium of the beam; and b) the 44

55 4.4 BEAM DYNAMICS Table 4.16: r.m.s. emittance increase between 3 MeV and 180 MeV for initially uniform and Gaussian beams Plane DTL CCDTL SCL Total [%] [%] [%] [%] x 5: Gaussian y 4: z 4: x 1: Uniform y z r.m.s. beam parameters have to be matched at each transition to avoid mismatching, which would induce emittance growth. The evolution of the phase advances per metre are shown in Fig (a). The focusing lattice is FDO. phase advance [ deg/m ] longitudinal transverse TraceWin - CEA/DSM/DAPNIA/SACM tune depression TraceWin - CEA/DSM/DAPNIA/SACM transverse longitudinal position [ m ] position [ m ] Fig. 4.31: (a) Transverse and longitudinal zerocurrent phase advance per metre (b) Transverse and longitudinal tune depression for 65 ma A Gaussian distribution cut at 4 with a transverse normalized r.m.s. emittance of 0.32 mm mrad and a longitudinal normalized r.m.s. emittance of 0.47 mm mrad is used at the entrance of the superconducting linac. The evolution of the beam is simulated from 180 MeV to 3.5 GeV using macro-particles. The peak current is 65 ma for a bunch frequency of MHz. No significant r.m.s. emittance growth is observed in the simulations. Figure 4.32 shows the evolution of the 4- envelopes End-to-end simulations End-to-end simulations have been carried out from 95 kev to 3.5 GeV. As the effect of beam neutralization on the low-energy end is still not completely understood, the end-to-end simulations start with a generated matched beam at the RFQ input. The beam contains macro-particles carrying a current of 70 ma, at an energy of 95 kev with an r.m.s. relative energy spread of 0.5%, a value that we expect from the source extraction voltage jitter. The transverse emittance is assumed to be 0.25 mm mrad (normalized, r.m.s.), with the beam being symmetrical in the two planes. Two distributions have been generated: 4D waterbag and Gaussian. We expect the behaviour of the beam to be bound between the evolution of these two distributions. The results of the tracking, without errors, are reported in Tables 4.17 and

56 4 DESIGN Fig. 4.32: 4- beam envelopes in the horizontal (x) and longitudinal (phase) plane Table 4.17: r.m.s. emittance growth and losses for an initially uniform distribution (end-to-end simulation) RFQ Chopper DTL CCDTL SCL SC Total Output energy [MeV] " x growth [%] " y growth [%] " z growth [%] Transmission [%] Length [m] Table 4.18: r.m.s. emittance growth and losses for an initially Gaussian distribution (end-to-end simulation) RFQ Chopper DTL CCDTL SCL SC Total Output energy [MeV] " x growth [%] " y growth [%] " z growth [%] Transmission [%] Length [m]

57 4.4 BEAM DYNAMICS Error studies The gradient and alignment errors quoted in the following are based partly on the experience at CERN and partly on already measured results of prototypes. They also correspond well to the error margins used for the SNS project [15] Method Two families of errors have to be considered. Static errors: the effect of these errors can be detected and cured with appropriate diagnostics and correctors. For example, beam position measurement coupled with steerers can compensate for quadrupole or cavity misalignments. The correction strategy should be known to be able to estimate the impact of static errors on beam dynamics. The effect of the static errors depends on the control system. Dynamic errors: the effect of these errors is assumed to be uncorrected. Fortunately, they usually have lower amplitude than static errors. There are, for example, the vibrations or the RF field variations (in phase or amplitude). They are responsible for beam oscillations around the corrected orbit. For both static and dynamic errors a uniform distribution was chosen with estimated maximum error amplitudes of A. The r.m.s. value is then given by A= p 3. If the real distribution of errors turns out not to be uniform, one should consider the study valid for the equivalent r.m.s. value. For reasons related purely to the availability and the location of manpower, the error study has been made for the normal-conducting and superconducting sections independently. The outcome of the error study of the normal-conducting part has been taken as an input beam jitter for the superconducting part. The results of each part (normal-conducting and superconducting) are reported in the following Normal-conducting section The error study for the normal-conducting part has been done by tracking a Gaussian distribution, containing particles over 1000 different linacs with random errors uniformly distributed within the values given in Table No correction scheme has been implemented. The results are reported in Table Table 4.19: Total error amplitudes used in the normal-conducting section Quadrupole gradient Quadrupole displacement Quadrupole rotations (x,y) Quadrupole rotations (z) Cavity field phase Cavity field amplitude 0:5% 0:1 mm 0:5 deg 0:2 deg 1:0 deg 1:0% Table 4.20: Sensitivity to errors in the normal-conducting section, average r.m.s. emittance increase, r.m.s. energy jitter assuming 4 MW beam power " x =" x " y =" y " z =" z E Losses av. [%] av. [%] av. [%] r.m.s. [kev] r.m.s. [deg] [W] All errors :

58 4 DESIGN Superconducting section The transport of 10 5 macro-particles has been simulated for each of 500 different linacs in order to compute statistics and to get a cumulative distribution of the beam which is represented by more than macro-particles. For each linac, all the errors, whose amplitudes are listed in Table 4.21, are combined. In a first stage, no input beam error is assumed. Table 4.21: Total error amplitudes used in the superconducting section Static Dynamic Quadrupole gradient [%] 0:5 0:05 Quadrupole displacement [mm] 0:5 0:01 Quadrupole rotations (x,y) [deg] 0:25 0:005 Quadrupole rotations (z) [deg] 0:5 0:05 BPM accuracy [mm] 0:1 Cavity displacement [mm] 0:5 0:01 Cavity rotations (x,y) [deg] 0:08 0:005 Cavity field phase [deg] 1:0 Cavity field amplitude [%] 1:0 For the transverse alignment, two (horizontal and vertical) steerers are located in the second quadrupole of each doublet. They are associated with two Beam Position Monitors (BPMs), located in the middle of the following doublet, which measure the beam centre in both planes. This scheme is repeated for each period. No longitudinal correction has been assumed. Assuming a normal-conducting front-end without errors one finds that: (a) the r.m.s. residual orbit along the machine due to dynamic errors and BPM imperfections (misalignments and measurement errors) is limited to 0.28 mm; (b) no losses are observed; (c) the r.m.s. emittance growth, relative to a linac without errors, amounts to an average of 27%, 26%, and 6% in the x, y, and z planes, respectively; (d) the r.m.s. output energy and phase jitter are 1.24 MeV and 0:52 ı, respectively. The above study has been repeated including a beam energy and phase jitter at the linac input. This is supposed to simulate the residual longitudinal errors from the normal-conducting part. Different error sets have been simulated to establish a limit for the jitter coming from the normal-conducting part. For each of the 500 linacs, the jitter values are randomly set between k E m and k m with E m and m representing the maximum energy and phase jitter amplitudes at the SCL output, respectively. They have been assumed as E m D 1:16 MeV (maximum, uniform distribution) and m D 12:6 ı (maximum, uniform distribution) at MHz. These values turned out to be too pessimistic and thus the parameter k is used to reduce them. The results are summarized in Table In order to avoid any beam loss in the superconducting part of the linac and assuming the errors of Table 4.21, the maximum remnant energy and phase errors at the output of the normal-conducting parts should be lower than 380 kev and 4 ı, respectively, for a uniform distribution. The equivalent r.m.s. values should be lower than 220 kev and 2:4 ı (352 MHz), which is more or less fulfilled by the simulation results for the normal-conducting section in Table Assuming these limits, the simulations of the superconducting section predict zero beam loss when combining transverse and longitudinal errors. The corresponding energy and phase jitter in this case amount to r.m.s. values of 0.65 ı and 1.2 MeV. 48

59 4.5 AUXILIARY SYSTEMS Table 4.22: r.m.s. emittance growth and losses in the superconducting section due to statistical errors for different values of input energy and phase jitter assuming 4 MW beam power k E input, max [kev] input, max [deg] " x growth [%] " y growth [%] " z growth [%] Average losses [W] Auxiliary systems Transfer line At the end of the superconducting linac section, the bunches are about 15 ps long and have an energy spread of the order of 4 MeV. To reduce space-charge effects and energy jitter at the accumulator injection, the bunch length has to be increased in the transfer line by at least a factor of 10, and the energy spread reduced by the same factor. This is achieved by two bunch rotation systems made of 700 MHz ˇ D 1 elliptical cavities identical to the ones used in the last section of the superconducting linac. The voltage to be applied for stretching the bunch in phase at the end of the linac depends on the length of the transfer line and the energy acceptance of the accumulator RF system. For the time being we assume injection into the ISR tunnel as detailed in CDR1 [3]. For this scenario, the first bunch rotation system (16 cavities, two cryo-modules) is placed at the linac exit to increase the energy spread in such a way that in the following long drift the bunch length is stretched up to the required value. The second bunch rotation section (two cavities) reduces the energy spread for injection into the RF bucket of the accumulator ring. The voltage on the bunch rotation cavities and their relative position depends on the longitudinal beam emittance at the linac exit. In the final layout, it can be re-adjusted for an effective longitudinal emittance that takes into account the effect of energy and phase jitter (Section 4.4.8). The first bunch rotation section, 29.4 m long, will be placed inside the linac tunnel. The second bunch rotator will be placed after 270 m. The general layout of this line is described in Table The beam dynamics is calculated starting with a distribution of particles at the entrance of the superconducting linac section. Phase and energy spread are presented in Fig and the corresponding phase-space distributions are shown in Fig Table 4.23: General layout of the transfer line Element Length [m] No. of cavities No. of focusing periods Cavity parameters Bunch rotation MV/m Drift Bunch rotation MV/m Total As an alternative, we have designed a transfer line which does not make use of the first bunch rotation but lets the beam de-bunch under the effect of the initial energy spread. This approach results in a longer line but saves 16 superconducting RF cavities. The line would then contain 48 focusing periods over a length of 730 m and only one five-cell cavity with an effective voltage of 11 MV for the final bunch rotation. For the time being there is space foreseen in the linac tunnel to accommodate two more cryo-modules for the active bunch rotation at the end of the linac. However, since it is likely that the location of the accumulator ring will change in future design revisions, the cavities and RF systems needed for the bunch rotation are not included in the parameter lists of this report. 49

60 4 DESIGN r.m.s. phase [deg] z [m] z [m] Fig. 4.33: Energy (left) and phase width (right) along the transfer line Fig. 4.34: Longitudinal phase space before (left) and after (right) the transfer line At 3.5 GeV the minimum arc radius to avoid H stripping is 220 m. This number assumes a dipole filling factor of 60% in the arc and a loss level due to H stripping of 0.1 W/m (compare Chapter 6, Section 6.1.2). The current understanding of H stripping due to black-body radiation will require cooling of the transfer-line beam pipes to 150 K (see also Chapter 6, Section 6.1.2) Pulsed operation and servo-system stability The studies for the CDR1 [3] addressed most of the principal problems arising from pulsed operation and remain valid for the present revised design of the SPL Lorentz detuning In CDR1 [3] it was planned to use cryostats, couplers, and other hardware from the decommissioned LEP superconducting RF system. The cavities were to be modified and fitted with stiffeners to enable medium-ˇ acceleration and operation in pulsed mode. At the time of publication of CDR1, the details of these stiffeners and with it the resulting stiffness along with the mechanical frequencies of the cavity tank tuner system (possibly weighted by the couplers) were not known. Since these are key parameters for the mechanical system behaviour, no detailed analysis concerning Lorentz detuning could be undertaken, relying on definite progress in the framework of TESLA [92], now absorbed in ILC [93], and SNS [15] which will very probably pave the way. The same argument holds for the present design. Once fully equipped prototype cavities really exist and are well characterized in terms of Lorentzforce detuning strength, frequency and mechanical Q-value of the principal mechanical mode(s) of the cavity tank tuner system, one can incorporate recent progress in the field of fast piezo-tuners [94] to reduce their effect. At this stage more detailed simulations have to be done. Possible modifications of the mechanical cavity design have to be undertaken if the simulations with all cavity errors and beam 50

61 4.5 AUXILIARY SYSTEMS show insufficient performance. For the time being, the developed computer code SPLinac [95], [96] considers only one principal mechanical mode since the LEP cavities were essentially detuning sensitive to the lowest longitudinal mechanical mode around 100 Hz. Should the new cavities have more than one essential mode, the program has to be upgraded for multiple mechanical modes; the same is true if the piezo-tuner corrections cannot be expressed by a simple scaling but show a time domain behaviour of their own. Then detailed measurements with artificial mechanical excitation and pulsed high RF fields have to be done. In the frame of the HIPPI Collaboration [50], two different ˇ D 0:47 elliptical cavities will be tested in pulsed mode in a horizontal cryostat at Saclay. At this low beta, the Lorentz-force detuning should be more severe than for the ˇ D 0:65 and ˇ D 1:0 cavities, which are foreseen for the SPL, and so the set-up, complete with couplers and tuners, should give a good estimate for the dynamical behaviour of the SPL cavities. The extra power needed to combat the effects of Lorentz-force detuning is estimated (from SNS) to be in the area of 30%. This additional power is only taken into account for the calculation of the total SPL power requirements, quoted in Table 5.9 and in the parameter list in the Appendix System errors and noise In CDR1 [3], simulations [97] were done with SPLinac based on a hypothetical set of test noise configurations. These can be created by vacuum pumps or cryogenic compressors (coherent), ground-movement and cryogenic liquid or gas movements (incoherent), and other sources. It should be noted that the mechanical conditions of the cavity tank tuner system combined with information on the noise sources (spectral density for noise, amplitude for discrete lines) are decisive for definite results. The simulations for CDR1 [3] showed that this effect is no major obstacle, hence the same can be expected for the present design. All RF error sources like scatter in the input coupling strength Q ext of the power couplers (e.g., observed in the LEP superconducting RF system), error distributions in the power splitters, and amplitude and phase errors in the RF vector sum have to be included in these simulations to determine how far the resulting beam quality is affected with regard to particle losses and longitudinal emittance growth Control instability for several pulsed cavities with RF vector-sum feedback Numerical simulations with SPLinac pointed to a control instability that can develop above a certain cavity field level for a pulsed RF system with Lorentz detuning sensitive cavities where more than one cavity is driven by the same high-power source and the vector sum of the cavity fields is fed back for control. A mathematical analysis [98] using an only slightly simplified model confirmed this chaotic effect and excludes a computer simulation artefact. From this analysis it became clear that the origin of this instability is the lack of control over all individual cavities since we have only a unique knob, the incident RF wave, but many degrees of freedom, one per cavity. Therefore even for a perfect RF vector sum the distribution of the work-load between cavities cannot be influenced and may diverge while respecting the sum. To remedy this situation and suppress the possibility for this instability, one knob per cavity becomes essential. A straightforward solution is the use of one transmitter per cavity, but the financial implications become an issue. In keeping a single transmitter for several cavities, one solution consists in the use of high-power phase and amplitude modulators [99] as already tested at CERN [100], [101] (Fig. 4.35), allowing the amplitude to be decreased and the phase of the incident wave to be turned towards an individual cavity in a limited way. The technical price to be paid is a permanent small over-power, finally wasted. In this set-up the main adaptations required to respect the vector sum are still done by the common transmitter, but the modulators compensate for individual cavity deviations from the average. 51

62 4 DESIGN Fig. 4.35: High-power tests of the amplitude and phase modulator, using a 300 kw/400 MHz LHC klystron and a sliding waveguide short circuit as load This choice has the additional advantage of also allowing the correction of the static errors in the power splitters and Q ext without having to create the possibility for local adjustment. Fast piezo-tuners will allow compensation of Lorentz-force detuning to a good degree. Once included in the RF chain, they will automatically reduce differences in tuning between cavities and can even be trimmed for this purpose with an additional controller. Therefore they can certainly take over part of the task of the above-mentioned phase and amplitude modulators. However, whether this compensation is sufficient to drop the phase and amplitude modulators completely, while leaving the remaining static errors unchanged, has to be explored by further R&D. 52